article single-cell profiling reveals an endothelium-mediated … · 2019. 7. 30. · article...

ARTICLE

Single-cell profiling reveals anendothelium-mediated immunomodulatory pathwayin the eye choroidGuillermo L. Lehmann1,2, Christin Hanke-Gogokhia1, Yang Hu3, Rohan Bareja3, Zelda Salfati1, Michael Ginsberg4, Daniel J. Nolan4,Santiago P. Mendez-Huergo5, Tomas Dalotto-Moreno5, Alexandre Wojcinski6, Francisca Ochoa2, Shemin Zeng7, Juan P. Cerliani5, Lampros Panagis2,Patrick J. Zager1, Robert F. Mullins7, Shuntaro Ogura8, Gerard A. Lutty8, Jakyung Bang9, Jonathan H. Zippin9, Carmelo Romano2,Gabriel A. Rabinovich5,10, Olivier Elemento3, Alexandra L. Joyner6, Shahin Rafii11, Enrique Rodriguez-Boulan1, and Ignacio Benedicto1,12

The activity and survival of retinal photoreceptors depend on support functions performed by the retinal pigment epithelium(RPE) and on oxygen and nutrients delivered by blood vessels in the underlying choroid. By combining single-cell and bulk RNAsequencing, we categorized mouse RPE/choroid cell types and characterized the tissue-specific transcriptomic features ofchoroidal endothelial cells. We found that choroidal endothelium adjacent to the RPE expresses high levels of IndianHedgehog and identified its downstream target as stromal GLI1+ mesenchymal stem cell–like cells. In vivo genetic impairmentof Hedgehog signaling induced significant loss of choroidal mast cells, as well as an altered inflammatory response andexacerbated visual function defects after retinal damage. Our studies reveal the cellular and molecular landscape of adultRPE/choroid and uncover a Hedgehog-regulated choroidal immunomodulatory signaling circuit. These results open newavenues for the study and treatment of retinal vascular diseases and choroid-related inflammatory blinding disorders.

IntroductionPhotoreceptors capture incoming photons and transform theminto electrical pulses that ultimately lead to visual perception.Their localization to the outer retina, distal to the incoming lightand visual processing neurons of the inner retina, is a successfulevolutionary design that allows vertebrate photoreceptors tobenefit from the critical support of two external eye layers, theretinal pigment epithelium (RPE) and the choroidal blood ves-sels. The RPE enables photoreceptor function by eliminatingstray light, phagocytosing photoreceptor waste products neces-sary for their renewal, recycling visual cycle components, andconstituting an essential component of the outer blood–retinalbarrier between choroidal circulation and the neural retina(Strauss, 2005). Whereas separate retinal blood vessels nourishthe inner retina, blood supplied by the choroidal circulation isthe main source of oxygen and nutrients for RPE and photo-receptors and the main evacuation route for retinal waste

(Nickla and Wallman, 2010). Thus, choroidal perfusion is es-sential for RPE and retinal homeostasis. However, this is likelynot the only role of choroidal endothelial cells (ECs), as recentstudies have shown that microvascular ECs are not passiveconduits for delivering blood but rather organ-specific factoriesof highly specialized sets of angiocrine factors that regulatetissue homeostasis and regeneration (Rafii et al., 2016). Recently,we provided initial evidence for this scenario in the eye. Wereported that developing and adult mouse choroidal ECs exhibitdifferent transcriptomes, and such transition is likely requiredfor the coordinated establishment of the outer blood–retinalbarrier and the acquisition of visual function (Benedicto et al.,2017). However, specific transcriptome features of choroidal ECscompared with retinal and extraocular ECs and the molecularidentities of nonendothelial choroidal cell types have not beendetermined in detail so far. RPE and choroid act as a functional

.............................................................................................................................................................................1Department of Ophthalmology, Margaret Dyson Vision Research Institute, Weill Cornell Medicine, New York, NY; 2Regeneron Pharmaceuticals, Inc., Tarrytown, NY;3Caryl and Israel Englander Institute for Precision Medicine, Department of Physiology and Biophysics, Weill Cornell Medicine, New York, NY; 4Angiocrine Bioscience, Inc.,San Diego, CA; 5Laboratorio de Inmunopatologıa, Instituto de Biologıa y Medicina Experimental, Consejo Nacional de Investigaciones Cientıficas y Tecnicas, Buenos Aires,Argentina; 6Developmental Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY; 7The University of Iowa Institute for Vision Research andDepartment of Ophthalmology and Visual Sciences, The University of Iowa, Iowa City, IA; 8Wilmer Ophthalmological Institute, Johns Hopkins Hospital, Baltimore, MD;9Department of Dermatology, Weill Cornell Medicine and New York-Presbyterian Hospital, New York, NY; 10Departamento de Quımica Biológica, Facultad de CienciasExactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina; 11Ansary Stem Cell Institute, Department of Medicine, Division of Regenerative Medicine, WeillCornell Medicine, New York, NY; 12Centro Nacional de Investigaciones Cardiovasculares (CNIC), Madrid, Spain.

Correspondence to Ignacio Benedicto: [email protected]; Enrique Rodriguez-Boulan: [email protected]; C. Hanke-Gogokhia’s present address is Duke Eye Center,Duke University Hospital, Durham, NC.

© 2020 Lehmann et al. This article is available under a Creative Commons License (Attribution 4.0 International, as described at https://creativecommons.org/licenses/by/4.0/).

Rockefeller University Press https://doi.org/10.1084/jem.20190730 1 of 21

J. Exp. Med. 2020 Vol. 217 No. 6 e20190730

Dow

nloaded from http://rupress.org/jem

/article-pdf/217/6/e20190730/1174063/jem_20190730.pdf by guest on 26 August 2021

https://orcid.org/0000-0002-6348-874X

https://orcid.org/0000-0002-5182-9594

https://orcid.org/0000-0002-5191-8933

https://orcid.org/0000-0002-5882-0189

https://orcid.org/0000-0002-0947-8735

https://orcid.org/0000-0001-7090-9605

https://orcid.org/0000-0003-2033-6233

https://orcid.org/0000-0001-8081-1847

mailto:[email protected]

mailto:[email protected]

https://creativecommons.org/licenses/by/4.0/

https://doi.org/10.1084/jem.20190730

http://crossmark.crossref.org/dialog/?doi=10.1084/jem.20190730&domain=pdf

unit, and defects in any of their supportive roles can causeretinal degeneration and blinding diseases such as age-relatedmacular degeneration (AMD). AMD, an incurable disease thataffects 8.7% of the worldwide population and 25% people >80 yrold, is characterized by photoreceptor loss secondary to dys-function or death of the RPE and choroidal ECs (Ambati andFowler, 2012; Mullins et al., 2011; Seddon et al., 2016; Wonget al., 2014). The etiology of AMD remains largely unknown,in part due to the very limited availability of molecular infor-mation about the different cell types that populate the choroidand the intercellular networks that maintain RPE/choroid tissuein a healthy and functional state.

Single-cell RNA sequencing (scRNAseq) is a recently devel-oped approach that allows efficient transcriptional profiling ofindividual cells in a given tissue or other complex cellularpopulations (Macosko et al., 2015). This technology has openedthe door to answering longstanding biological questions on tis-sue cell diversity, heterogeneity of cellular responses, and reg-ulatory signaling networks that would be difficult or impossibleto answer using traditional physiological or biochemical ap-proaches. A recent landmark study characterized by scRNAseqthe molecular identity of all cell types in the mouse neural retinabut did not analyze RPE or choroid (Macosko et al., 2015). Morerecently, scRNAseq was used to create a mouse cell atlas cov-ering all major organs, but RPE and choroid were not included(Han et al., 2018). Here, we report a scRNAseq analysis of adultmouse RPE/choroid and provide the transcriptional signature ofchoroidal ECs compared with other tissue-specific ECs, includ-ing retinal ECs. Our studies uncover the molecular identity ofthe major choroidal cell types, including three different ECsubtypes and a previously uncharacterized population of me-senchymal stem cell (MSC)–like cells. Moreover, MSC-like cellsare the target of EC-secreted Indian Hedgehog (IHH), which wasfound to be a critical modulator of choroidal and retinal in-flammatory responses. These results open new avenues for thestudy and treatment of retinal vascular diseases and choroid-related inflammatory blinding disorders such as AMD.

ResultsCharacterization of mouse RPE/choroid by scRNAseqTranscriptional profiling at single-cell resolution requires high-quality cell suspensions. To this end, we developed a fast andefficient procedure to isolate single cells from RPE/choroid bysequential tissue digestion and cell sorting. We used FACS tooptimize the capture of single, viable cells (TO-PRO-3 negative)and separate nucleated cells (Hoechst positive) from cell debris(Figs. 1 a and Fig. S1 a). We surgically dissected RPE/choroidtissue from eyes enucleated from 90-d-old mice of two differentstrains, C57BL/6J and B6129PF1/J, and sequenced 3,996 and 3,727cells from each strain, respectively. To avoid any potential sex-related bias during our analyses, we pooled tissue from two eyes,one male and one female, for each strain during enzymatic di-gestion. In addition, this approach allowed us to estimate that∼4% of isolated cells were present as doublets (i.e., two cellsbarcodedwith the same sequence), based on the assumption thatmale–female doublets accounted for 50% of total doublets (Fig.

S1 b; Macosko et al., 2015). Unsupervised clustering of individualcell transcriptomes based on similarity in overall gene expres-sion identified 13 transcriptionally distinct clusters within RPE/choroid cells (Fig. 1 b), with remarkably similar results for bothstrains (Fig. S1 c). Using the FindConservedMarkers function inSeurat, we found ≥65 conserved cluster-specific genes for eachcluster in both mouse strains (adjusted [adj] P < 0.05; Data S1 a).Based on these gene lists and the expression of known cell-type–specific markers, we categorized these clusters as RPE,ECs, stromal cells, smooth muscle cells, melanocytes, hemato-poietic cells, and Schwann cells (Fig. 1 c, B6129PF1/J strain; Fig.S1 d, C57BL/6J strain; and Table 1). This analysis identifiedsubtypes within ECs (clusters 2–4), stromal cells (clusters 5–8),and Schwann cells (clusters 12 and 13). The complete datasets forboth strains are reported in Data S1 (b and c).

Identification of transcriptionally distinct choroidalEC subtypesThe choroidal vasculature plays key roles in retinal homeostasisand in the pathogenesis of blinding diseases such as AMD(Bhutto and Lutty, 2012; Nickla and Wallman, 2010; Whitmoreet al., 2015); hence, we were particularly interested in furthercharacterizing choroidal EC subtypes. EC clusters 2–4 presentedsimilar expression levels of the generic EC markers Cdh5 (VE-cadherin), Pecam1 (CD31), Flt1 (VEGFR1), and Cldn5 (claudin-5);however, they displayed remarkable variations in the expres-sion of other markers. Because Kdr (VEGFR2) was expressed atdecreasing levels in EC clusters 2–4 (Fig. 1 c and Fig. S1 d), wetermed these clusters Kdrhigh, Kdrmed, and Kdrlow ECs, respec-tively. To study the most significant transcriptional differencesbetween Kdrhigh and Kdrlow ECs, we selected the genes whoseexpression in Kdr high was detected by an average of ≥1 normal-ized unique molecular identifier (UMI), and the average numberof normalized UMIs was at least fivefold higher than in Kdrlow

ECs and vice versa. Average normalized UMIs for each gene inKdrmed were also included in the analysis for comparison pur-poses (Data S1, d and e). Using this filtering approach, we found19 genes specifically enriched in Kdrhigh compared with Kdrlow

(Fig. 1 d, top panel) and 19 genes with the opposite expressionpattern (Fig. 1 d, bottom panel). In general, gene expression inKdrmed was intermediate between Kdrhigh and Kdrlow, suggestingthat Kdrmed constitutes a transitional EC population between theother two.

The choriocapillaris is the capillary network immediatelyadjacent to the RPE basement membrane that provides nutritionand a waste evacuation route to the outer retina, and it ischaracterized histologically by the presence of fenestrations inthe EC plasma membrane (Lutty et al., 2010; Nickla andWallman, 2010). As the gene with highest expression in Kdrhigh

ECs is Plvap (Fig. 1 d, top panel), which encodes the fenestrae-associated protein PV1 (Stan et al., 2012), it is likely that theKdrhigh cluster represents choriocapillaris ECs, whereas theKdrlow cluster corresponds to ECs from deeper choroid capillar-ies. To test this hypothesis, we performed immunofluorescenceexperiments using antibodies against KDR and RPE65, an RPEmarker. KDR was detected in the innermost choroidal layer veryclose to the RPE (Fig. 1 e, left), consistent with the location of the

Lehmann et al. Journal of Experimental Medicine 2 of 21

Endothelial immunomodulation in the eye choroid https://doi.org/10.1084/jem.20190730

Dow



https://doi.org/10.1084/jem.20190730

Figure 1. Characterization of mouse RPE/choroid by scRNAseq and identification of transcriptionally distinct choroidal EC subtypes. (a) Scheme ofthe preparation of RPE/choroid single-cell suspensions by sequential enzymatic digestion and cell sorting. Hoechst+/TO-PRO-3− live cells were subjected toscRNAseq. (b) Principal-component analysis of 7,723 single-cell expression profiles obtained from RPE/choroid tissue from four mice (C57BL/6J and B6129PF1/Jstrains, one male and one female per strain). Data are shown in two dimensions using tSNE. Unsupervised analysis clustered cells into 13 transcriptionallydistinct cell populations, each plotted in a different color. (c) Identification of cell types in RPE/choroid tissue from B6129PF1/J mice according to the averageexpression of known markers in each cluster. Cluster number colors are the same as in panel b. Relative gene expression among cell types was calculated usingthe average normalized UMIs in each cluster and represented as the percentage of the cluster with maximum expression. White, 0%; red, 100%. (d) Genes withat least fivefold differential expression between Kdrhigh (green bars) and Kdrlow (purple bars) ECs (Benjamini–Hochberg adjusted likelihood-ratio test, adj P <0.05). Only genes with maximum average normalized UMIs ≥1 are shown. Relative gene expression in Kdrmed ECs (brown bars) is also included. Top: Kdrhigh

expression at least fivefold higher than Kdrlow. Bottom, Kdrlow expression at least fivefold higher than Kdrhigh. (e) Immunofluorescence analysis of KDR (red),RPE65 (RPE marker), and claudin-5 (pan-endothelial marker; green) expression in mouse eye cryosections. Nuclear staining with DAPI (blue) and differentialinterference contrast images are also included in merged panels. Position of RPE and choroid (Ch) is indicated on the left of merged images. Scale bars, 5 µm(left) and 10 µm (right). Images are representative of four mice.



Dow



https://doi.org/10.1084/jem.20190730

choriocapillaris. We also costained KDR and claudin-5, a pan-endothelial marker expressed at similar levels in Kdrhigh, Kdrmed,and Kdrlow choroidal ECs (Fig. 1 c and Fig. S1 d). Claudin-5 ex-pression was found both in KDR+ ECs and in deeper choroidallayers closer to the sclera with undetectable KDR expression(Fig. 1 e, right), the latter likely corresponding to Kdrlow ECs.These experiments strongly suggest that Kdrhigh and Kdrlow

clusters correspond to transcriptionally and geographically dis-tinct populations of choroidal ECs.

Tissue-specific enrichment of Ihh expression in choriocapillarisECsIt is becoming evident that adult ECs have tissue-specific tran-scriptomes reflecting context-dependent physiological and re-generative roles (Rafii et al., 2016). We recently reported thetranscriptome of developing and adult mouse choroidal ECs,isolated by intravital staining of the EC marker VE-cadherin,followed by RPE/choroid digestion, flow cytometry sorting, andbulk RNA sequencing (RNAseq; Benedicto et al., 2017). Here, wefollowed the same approach to compare the transcriptionalprofile of ECs from adult mouse choroid, neural retina, lung,heart, and liver (Fig. 2 a). Hierarchical clustering analysisdemonstrated marked transcriptome differences between thevarious EC types (Fig. 2 b). We identified a list of 99 “signaturegenes” of mouse choroidal ECs that were expressed at levels atleast fivefold higher than in the other ECs, with a detectionthreshold set at ≥1 fragments per kilobase per million reads(FPKMs; Data S1 f). By comparing our bulk and single cellRNAseq data, we were able to determine the relative expressionof every choroidal EC signature gene relative to other RPE/

choroid cell types. Out of 99 genes, 21 were specifically enrichedin choroidal ECs (Fig. 2 c; >50% of the sum of average normalizedUMIs from all RPE/choroid cell types), from which 16 weresignificantly more abundant (adj P < 0.05) in Kdrhigh (chorio-capillaris) ECs than in Kdrmed and Kdrlow choroidal ECs (Data S1, dand g). Our studies provide the first comprehensive list of spe-cific molecular markers for all choroidal ECs relative to not onlyother choroidal cell types but also other tissue-specific ECs. Theydemonstrate that the choriocapillaris is the most specializedvascular subtype within the choroid and provide an importantplatform for future studies on choroid physiopathology and thepathogenesis of choroid-based blinding diseases.

One of the genes specifically enriched in Kdrhigh (choriocap-illaris) ECs was Ihh (Fig. 2 c). The three secreted Hedgehog (HH)proteins (IHH, Sonic Hedgehog, and Desert Hedgehog) act in aparacrine manner, and in most tissues are expressed by epi-thelial cells and signal to the surrounding mesenchyme (Petrovaand Joyner, 2014; Wu et al., 2017). Detailed examination of ourbulk RNAseq data revealed that Ihh expression in choroidal ECswas >340-fold higher than in the rest of the tissue-specific ECs(Fig. 2 d), an observation that was confirmed by real-time PCRassays (Fig. 2 e). Expression of the generic EC markers Cdh5,Pecam1, and Kdr was similar among all tissue-specific ECs, fur-ther highlighting the specificity of Ihh enrichment (Fig. 2 d).Importantly, IHH expression in human RPE/choroid tissue fromhealthy donors was markedly higher than in neural retina, lung,liver, and heart, as assessed by real-time PCR (Fig. 2 f). Ourobservation that Ihh is expressed at high levels in adult chorio-capillaris is particularly intriguing, as IHH is known to regulateprenatal mouse eye development (Dakubo et al., 2008).

Table 1. Mouse RPE/choroid cell types identified by scRNAseq

Clusternumber

Cell type Cell number Percentage of total Average UMIs/cell Average genes/cell Average UMIs/gene

C57BL/6J mice

B6129PF1/Jmice

C57BL/6J mice

B6129PF1/Jmice

C57BL/6J mice

B6129PF1/Jmice

C57BL/6J mice

B6129PF1/Jmice

C57BL/6J mice

B6129PF1/Jmice

1 RPE 402 445 10.1 11.9 4,860 7,514 1,468 1,911 3.02 3.51

2 ECs 405 366 10.1 9.8 4,550 5,151 1,920 2,090 2.18 2.32

3 ECs 550 435 13.8 11.7 4,847 5,570 1,943 2,139 2.28 2.42

4 ECs 74 63 1.9 1.7 4,749 5,196 1,894 2,025 2.26 2.37

5 Stromal cells 776 797 19.4 21.4 3,255 3,723 1,524 1,645 2.09 2.21

6 Stromal cells 610 552 15.3 14.8 3,171 3,796 1,491 1,681 2.07 2.21

7 Stromal cells 223 229 5.6 6.1 3,695 4,518 1,684 1,894 2.14 2.32

8 Stromal cells 158 166 4.0 4.5 2,995 3,944 1,391 1,661 2.13 2.33

9 Smooth musclecells

476 400 11.9 10.7 3,863 4,769 1,583 1,814 2.38 2.58

10 Melanocytes 28 20 0.7 0.5 3,913 5,008 1,621 1,847 2.32 2.59

11 Hematopoieticcells

140 116 3.5 3.1 4,573 4,960 1,528 1,496 2.83 3.19

12 Schwann cells 108 106 2.7 2.8 2,525 2,835 1,306 1,418 1.90 1.96

13 Schwann cells 46 32 1.2 0.9 2,606 3,532 1,227 1,495 2.08 2.31

Total 3,996 3,727

Data are from C57BL/6J and B6129PF1/J mice (pooled cells from one male and one female for each strain).



Dow



https://doi.org/10.1084/jem.20190730

Figure 2. Transcriptional analysis of tissue-specific mouse ECs and Ihh expression in choriocapillaris ECs. (a) Diagram showing the isolation of tissue-specific ECs after intravital EC labeling. For RPE/choroid and neural retina, seven animals (14 eyes) were used per isolation; for lung, liver, and heart, one animalwas used per isolation. Bulk RNAseq was performed using RNA from three independent isolations. (b) Hierarchical clustering of FPKM profiles showingseparate clustering of tissue-specific ECs (n = 3). (c) Signature genes of mouse choroidal ECs and relative expression among RPE/choroid cell types. Using datafrom the bulk RNAseq analyses of tissue-specific ECs, the list was assembled by selecting the genes with expression levels at least fivefold higher in choroidalECs compared with the rest of the tissue-specific ECs (n = 3, Benjamini–Hochberg corrected adj P < 0.05), with a detection threshold set at ≥1 FPKM. The listwas then interrogated against our scRNAseq data to assess the relative expression of each gene among all cell types, which was calculated using the averagenormalized UMIs in each cluster and represented as the percentage of the cluster with maximum expression. White, 0%; red, 100%. The column labeled “CCexpression” shows the average normalized UMIs (log10) for each gene in Kdrhigh ECs, i.e., the choriocapillaris (CC), represented in a white-purple scale. Black-gray bars on the right represent EC specificity compared with the rest of RPE/choroid cell types. For each gene, the sum of the average normalized UMIs inKdrhigh, Kdrmed, and Kdrlow ECs was represented as the percentage of total average normalized UMIs (the sum of average normalized UMIs in all clusters). Theblack portion of the bars represents the percentage of estimated EC specificity. Only genes with >50% EC specificity are shown. (d) Bulk RNAseq resultsshowing the expression (FPKM) of Ihh (red) and the EC markers Cdh5 (black), Pecam1 (gray), and Kdr (blue) in tissue-specific ECs (n = 3). ***, Benjamini–Hochberg corrected adj P < 0.001 all groups versus choroidal ECs. (e) Real-time PCR showing relative Ihh expression in tissue-specific ECs. Results arepresented as the percentage of the average value for choroidal ECs (n = 3, ANOVA + Bonferroni test). ***, P < 0.001 all groups versus choroidal ECs. nd, notdetected. (f) Real-time PCR showing relative IHH expression in human whole tissues (RPE/choroid, n = 4; neural retina, n = 5; commercially obtained lung, liver,and heart, n = 1). Biological replicates are shown as black dots. Red lines show the average expression in RPE/choroid and neural retina. Results are presentedas the percentage of the average value for RPE/choroid tissue.



Dow



https://doi.org/10.1084/jem.20190730

However, its expression and potential role in the adult mouseeye has never been studied before. These results prompted usto study the signaling pathway downstream of choroidal EC-expressed Ihh.

Localization and characterization of choroidal GLI1+ IHH targetcellsBecause IHH is a secreted protein expected to signal to thesurrounding mesenchymal cells, we sought to identify cell typeswithin RPE/choroid capable of responding to EC-expressed Ihh.Visualization of Ihh-expressing cells on t-distributed stochasticneighbor embedding (tSNE) plots confirmed that Ihh transcriptsare constitutively enriched in the Kdrhigh subcluster of ECs, whilethe expression of the gene encoding the transcriptional activatorGLI1, a transcriptional target dependent on HH signaling andthus a readout of canonical HH pathway activity (Ahn andJoyner, 2005; Bai et al., 2004; Petrova and Joyner, 2014; Wuet al., 2017), was specifically increased in stromal cell clusters5–8 relative to other choroidal cell types (Fig. 3 a). More detailedanalysis of our scRNAseq data confirmed the stromal specificityof Gli1 expression (Fig. 3 b). Although the Ptch1 gene encoding theHH receptor presented a broader expression pattern than Gli1, itwas clearly enriched in stromal cells (Fig. 3 b). This is consistentwith the fact that Ptch1, in addition to encoding the HH receptor,is up-regulated in response to HH signaling (Petrova and Joyner,2014; Wu et al., 2017). To spatially localize choroidal GLI1+ cells,we used albino Gli1GFP/+ reporter mice in which GFP expressionis controlled by the endogenous Gli1 promoter of one of the al-leles (Brownell et al., 2011). Thus, HH-responding cells can bereadily identified as GFP+ cells. Immunostaining analysis ofchorioretinal sections from Gli1GFP/+ mice showed that GFP+ cellswere primarily localized surrounding VE-cadherin+ ECs andcolocalized with the stromal marker PDGFRβ (Fig. 3 c). Theseresults indicate that the main paracrine targets of EC-secretedIHH are perivascular stromal cells.

Previous work demonstrated the constitutive presence ofperivascular GLI1+ MSC-like cells in several adult mouse tissues.These cells represent between 0.01% and 0.17% of total tissuecells and are capable of differentiating into adipocytes, osteo-blasts, and chondrocytes in vitro (Kramann et al., 2015). To as-sess their potential similarity with choroidal GLI1+ cellsdescribed here, we first studied the percentage of GFP+ cells inwhole digested tissue by flow cytometry. We found that thepopulation of resident GFP+ cells in RPE/choroid was 15.8 ± 1.3%,a much higher percentage compared with lung tissue (1.28 ±0.2%; Fig. 3 d and Fig. S2). Next, we phenotyped choroidal GFP+

cells by flow cytometry studying the expression of a repertoireof cell surface markers previously used to characterize GLI1+

MSC-like cells in other tissues (Kramann et al., 2015). We foundthat virtually all choroidal GFP+ cells were positive for PDGFRβand CD29 and negative for the hematopoietic lineage markerCD45 and the EC marker CD31, and ∼50% expressed CD105 andSCA1 (Fig. 3 e). To study whether choroidal GLI1+ cells are plu-ripotent in vitro, we isolated GLI1+ cells from RPE/choroid ofGli1GFP/+ mice by cell sorting based on GFP expression. IsolatedGFP+ cells were plastic adherent and could be differentiated intoosteoblast- and adipocyte-like cells (Fig. 3, f and g). Since it is

well established that MSCs inhibit T cell proliferation in vitro(Uccelli et al., 2008), we tested whether choroidal GLI1+ cellsexert the same effect. Mouse splenocytes were labeled withCFSE, polyclonally activated, and cultured in the absence orpresence of allogeneic choroidal GLI1+ cells. Proliferation of CD4+

and CD8+ T cells was inhibited in the presence of GLI1+ cells in adose-dependent manner (Fig. 3, h–j). Overall, our characteriza-tion of choroidal GLI1+ cells is in agreement with the MSC-likeidentity previously reported for GLI1+ perivascular cells fromother tissues (Kramann et al., 2015).

RPE/choroid transcriptional alterations after EC-specificdeletion of Ihh in adult miceTo directly study the role of choroidal EC-expressed Ihh inchoroidal and retinal homeostasis of adult mice, we generated amouse model that enabled EC-specific, inducible Ihh deletion.Cdh5-PAC-CreERT2 mice, in which a tamoxifen-inducible Cre isexpressed under the control of the EC-specific Cdh5 promoter(Wang et al., 2010), were crossed withmice harboring floxed Ihhto obtain Cdh5-PAC-CreERT2 IhhloxP/loxP mice (hereafter termedIhhiΔEC/iΔEC). To assess Ihh deletion efficiency in our induciblesystem, we treated control (Cdh5-PAC-CreERT2 Ihh+/+, hereaftertermed Ihh+/+) and IhhiΔEC/ΔiEC adult mice with tamoxifen andassessed Ihh mRNA levels in RPE/choroid by real-time PCR. Weobserved that IhhiΔEC/iΔEC mice presented a ∼70% reduction inIhh mRNA levels. Confirming that IHH signaling is diminishedin mutants, Gli1 expression was similarly decreased (Fig. 4 a).These results confirmed our scRNAseq data showing that themain source of Ihh expression in RPE/choroid tissue are cho-roidal ECs and, importantly, demonstrate that choroidal Gli1expression depends on EC-expressed Ihh.

Next, we performed bulk RNAseq analyses of RPE/choroidtissue obtained from tamoxifen-treated Ihh+/+ and IhhiΔEC/iΔEC

mice (Data S1 h). To analyze the impact of EC-specific Ihh de-letion on the different cell types within RPE/choroid tissue, weassembled a list of genes whose expression in Ihh+/+ mice was≥1 FPKM and at least twofold higher than in IhhiΔEC/iΔECmice (adjP < 0.05). In other words, we selected for genes highly expressedin RPE/choroid that were markedly down-regulated after EC-specific Ihh deletion. We then used our scRNAseq analyses ofwild-type mice to plot the relative expression of each of thesegenes in all RPE/choroid cell clusters (Fig. 4 b). We found thatrelative expression of these genes in RPE was very low. Thisresult indicated that, at least under basal conditions, EC-specificIhh deletion had little impact on RPE transcriptome comparedwith the rest of the cell types. Out of the 20 genes included in thelist, we found 8 that were clearly enriched in stromal cell clus-ters 5–8, including the HH-induced genes Gli1 and Hhip. Thisfinding was expected, given our previous observation thatstromal cells are the main choroidal Ihh target. Interestingly,EC-specific Ihh deletion not only altered genes expressed by HH-responding stromal cells but also down-regulated genes ex-pressed almost exclusively in other choroidal cell populationsthat are not direct targets of IHH. That was the case for Syt4,Cort, and Sgcg, which are highly enriched in melanocytes,and Gpr34 and Cma1, which are selectively detected in hemato-poietic cells. These results strongly suggest that deletion of



Dow



https://doi.org/10.1084/jem.20190730

Figure 3. Characterization of choroidal GLI1+ cells. (a) tSNE graphs showing the predominant expression of Ihh and Gli1 in KDRhigh ECs (CC, choriocapillaris)and stromal cells, respectively. (b) Heatmap showing the relative expression of Ihh, Gli1, and Ptch1 in RPE/choroid cell types, calculated using the averagenormalized UMIs in each cluster and represented as the percentage of the cluster with maximum expression. White, 0%; red, 100%. (c) Immunofluorescenceassays using eye cryosections from Gli1GFP/+ mice show the perivascular localization of choroidal GLI1+ cells (GFP+, green). Top: Localization of GFP+ cells andVE-cadherin+ ECs (purple) is mutually exclusive. Bottom: GFP+ cells colocalize with the stromal marker PDGFRβ (purple). Colocalization in the merged images isshown in white. Nuclei were stained with Hoechst (blue). Zoomed representative regions of the merged images are shown as insets on the right panels. ONL,



Dow



https://doi.org/10.1084/jem.20190730

EC-expressed Ihh indirectly affects choroidal melanocytes andhematopoietic cells. Finally, a group of genes (Jun, Socs3, Phlda3,Mcm6, and Arhgap33) presented homogeneous relative expres-sion inmany choroidal cell types. However, because clusters 5–8constitute ∼50% of the total cellular content of RPE/choroidtissue (Table 1), this analysis strategy is not appropriate to elu-cidate whether the decreased expression of these genes by atleast twofold in the RPE/choroid of IhhiΔEC/iΔECmice results fromtheir down-regulation in choroidal stromal cells, nonstromalcells, or both. Taken together, our bioinformatics analyses sug-gest that EC-expressed Ihh directly regulates gene expression inGLI1+ perivascular MSC-like cells, which may in turn inducealterations in other choroidal cell types, including melanocytesand hematopoietic cells.

Impairment of HH signaling causes loss of choroidal mast cellsCma1 encodes chymase 1, a protease specifically expressed bymast cells (Dwyer et al., 2016). Mast cells control innate andadaptive immune responses and trigger IgE-associated allergicinflammation (Galli et al., 2008). Although the existence ofchoroidal mast cells in humans and rodents is well established(May, 1999; McMenamin, 1997; McMenamin and Polla, 2013),their role in the choroid is unknown. The decrease of Cma1 ex-pression in IhhiΔEC/iΔEC mice (Fig. 4 b) suggested that EC-specificIhh expression may control the number and function of cho-roidal mast cells. We reanalyzed our RNAseq data in search forother mast cell specific markers (Dwyer et al., 2016) and foundthat in addition to Cma1 (adj P = 0.02), Mcpt4 and Tpsb2 ex-pression was also reduced in IhhiΔEC/iΔECmice more than twofold(adj P = 0.07 and 0.06, respectively; Data S1 h), albeit notreaching the statistical significance threshold of adj P < 0.05used to filter data shown in Fig. 4 b. Real-time PCR assays re-vealed significantly decreased expression of Cma1, Mcpt4, andTpsb2 in RPE/choroid tissue from IhhiΔEC/iΔEC mice (Fig. 4 c),suggesting that EC-specific Ihh deletion may result in reducednumbers of choroidal mast cells. To address this possibility, weperformed avidin staining assays to detect mast cells (Tharpet al., 1985; Veerappan et al., 2008) in RPE/choroid flatmounts.Because RPE and choroidal pigmentation interferes with flat-mount imaging, these experiments were performed using albinomice generated by breeding the C57BL/6J-derived IhhiΔEC/iΔEC

pigmented line with albino mice. Quantitative image analysisshowed reduced numbers of choroidal mast cells in albino miceafter EC-specific Ihh deletion (Fig. 4, d and e). Next, we testedwhether alteration of the HH pathway in GLI1+ cells also resultedin reduced numbers of choroidal mast cells. We intercrossedalbino Gli1GFP/+ animals to obtain Gli1GFP/GFP mice, which are in-deed null for Gli1 function. Although Gli1 is dispensable fornormal mouse development (Bai et al., 2002; Park et al., 2000),Gli1 deletion induces altered phenotypes in particular contexts(Lees et al., 2008; Merchant et al., 2010). Real-time PCR assaysshowed that expression of Cma1, Mcpt4, and Tpsb2 in RPE/cho-roid tissue from Gli1GFP/GFP mice was significantly reducedcompared with Gli1+/+ mice (Fig. S3 a). Avidin staining assaysshowed reduced numbers of choroidal mast cells in Gli1GFP/GFP

mice (Fig. S3, b and c), confirming the importance of HH sig-naling for choroidal mast cell fate. These results demonstratethat blunting HH signaling, either by decreasing ligand levels(IHH) or impairing target cell response (Gli1 deletion), results inreduced numbers of choroidal mast cells. One attractive expla-nation for this observation is that mast cell survival in thechoroid may rely on factors provided by HH-stimulated GLI1+

cells. As a proxy to test this hypothesis, we performed in vitroassays using RBL-2H3 cells, a cell line widely used to study somefeatures of mast cell physiology (Passante and Frankish, 2009).Fluorescently labeled RBL-2H3 cells were cultured alone or to-gether with GLI1+ cells, in the absence or presence of the HHpathway agonist Smoothened agonist (SAG; Chen et al., 2002).After 24 h of serum starvation, viability of nonadherent RBL-2H3 cells was evaluated by DAPI staining. The presence of GLI1+

cells reduced the percentage of dead RBL-2H3 cells (DAPI+),which was further decreased by the addition of SAG (Fig. 4, fand g). Importantly, SAG treatment in the absence of GLI1+ cellshad no effect on RBL-2H3 cell viability. In sum, our in vivo andin vitro assays support a model in which EC-secreted IHHstimulates choroidal GLI1+ cells to promote mast cell survival.

HH-stimulated GLI1+ cells induce melanocyte proliferationBecause reduced RPE/choroid expression of mast cell markerscorrelated well with the actual number of choroidal mast cells,we reasoned that reduced expression of the melanocyte-enriched genes Syt4, Cort, and Sgcg in the RPE/choroid of

outer nuclear layer; OS/IS, outer segments/inner segments. Scale bar, 20 µm. Results are representative of at least two independent experiments.(d) Percentage of GLI1+ cells (GFP+) in RPE/choroid and lung. Left: Representative quantification of choroidal GLI1+ cells from Gli1GFP/+ mice by flow cytometry(green lines). Gli1+/+mice (black lines) were used as a negative control to set background fluorescence levels. Right: Choroid and lung GLI1+ cells were quantifiedby flow cytometry and represented as the percentage of total cells in each tissue (n = 3, t test). ***, P < 0.001. (e) Characterization of choroidal GLI1+ cells byflow cytometry. Gated GLI1+ cells (GFP+) were analyzed for the expression of CD45 (hematopoietic marker), CD31 (EC marker), and the MSC markers CD105,CD29, PDGFRβ, and SCA1. Unstained samples (black lines) were used to set background fluorescence levels. Numbers show the percentage ± SD (n = 3) of GFP+

cells positive for the different markers. (f) Choroidal GLI1+ cells were isolated from Gli1GFP/+mice by cell sorting (GFP+ cells) and expanded in culture. GLI1+ cellswere exposed to adipogenic (Ad) or osteogenic (Os) media, and expression of Fabp4 and Spp1 (adipocyte and osteoblast markers, respectively) was assessed byreal-time PCR at days 0 and 21 of differentiation (n = 3, ANOVA + Bonferroni test). ***, P < 0.001. (g) GLI1+ cells were cultured in osteogenic (left) or adipogenic(right) media for 21 d. Osteogenic and adipogenic differentiation was assessed by the appearance of calcium deposits (Alizarin red S staining) or intracellularlipid vesicles (oil red O staining), respectively. Day 0 cultures were included as controls. (h) Inhibition of T cell proliferation by GLI1+ cells (experiment outline).CFSE-labeled allogeneic mouse splenocytes were activated with anti-CD3 antibody and cultured alone or in the presence of choroid GLI1+ cells at differentratios (GLI1+ cells/splenocytes, 1:1 and 1:4) for 4 d. (i) CFSE dilution histograms showing that the proliferation of CD4+ and CD8+ T cells (white histogram) isstrongly inhibited by choroidal GLI1+ cells at both 1:1 and 1:4 ratios. (j) Quantification of panel i using the division index (average number of divisions afterstimulation) and the percentage of divided cells (No GLI1+ cells, n = 2; 1:1, n = 3; 1:4, n = 3; ANOVA + Bonferroni test). *, P < 0.05; **, P < 0.01; ***, P < 0.001.Results in panels f–j are representative of two independent experiments.



Dow



https://doi.org/10.1084/jem.20190730

Figure 4. RPE/choroid transcriptional alterations and loss of choroidal mast cells after EC-specific Ihh deletion. (a) Real-time PCR showing Ihh and Gli1expression levels in RPE/choroid tissue from Ihh+/+ and IhhiΔEC/iΔECmice (black and blue dots, respectively). Individual dots correspond to different animals, andred bars represent average values. Relative gene expression is presented as the percentage of Ihh+/+ mice (n = 6, t test). **, P < 0.01. (b) RNAseq analysis ofRPE/choroid tissue from Ihh+/+ and IhhiΔEC/iΔEC mice and cross-comparison with scRNAseq data. The heatmap shows genes with ≥1 FPKM in Ihh+/+ mice thatwere significantly down-regulated in IhhiΔEC/iΔEC mice (n = 3, at least twofold change, Benjamini–Hochberg corrected adj P < 0.05). Their relative expressionacross RPE/choroid cell types was calculated using the average normalized UMIs in each cluster and represented as the percentage of the cluster withmaximum expression. White, 0%; red, 100%. (c) Real-time PCR showing the expression of mast cell markers in RPE/choroid tissue from Ihh+/+ and IhhiΔEC/iΔEC

mice (n = 3, t test). *, P < 0.05; **, P < 0.01. Results are represented as in panel a. (d) Representative avidin staining assay to localize choroidal mast cells usingflatmounts from albino Ihh+/+ and IhhiΔEC/iΔEC mice. Scale bar, 1,000 µm. (e) Quantification of avidin staining assays shown in panel d. Results are expressed as



Dow



https://doi.org/10.1084/jem.20190730

IhhiΔEC/iΔEC mice (Fig. 4 b) may also indicate loss of choroidalmelanocytes. We used our scRNAseq data (Data S1 c) to compilea complete list of choroidal melanocyte-enriched genes (averagenormalized UMIs at least fivefold higher compared with othercell types; adj P < 0.05) and analyzed their expression in theRPE/choroid of Ihh+/+ and IhhiΔEC/iΔEC mice as determined bybulk RNAseq (Data S1 h). We found a trend of reduced expres-sion in IhhiΔEC/iΔEC mice (65 out of 70 melanocyte-enrichedgenes; Data S1 i), and all genes with significant differential ex-pression (adj P < 0.05) were down-regulated in IhhiΔEC/iΔEC mice(Fig. S3 d). Conversely, using the same approach, we observedthat Ihh+/+ and IhhiΔEC/iΔEC mice expressed similar levels of allsmooth muscle cell– and choriocapillaris-enriched genes (withthe exception of Ihh for the latter, as expected), suggesting thatthe reduction of melanocyte markers was specific (Data S1 i).These observations indicate that EC-specific Ihh deletion resultsin reduced numbers of choroidal melanocytes. To directly testwhether HH signaling modulates melanocyte proliferation, weperformed in vitro experiments using choroidal GLI1+ cells andthe mouse melanocyte cell line melan-a (Bennett et al., 1987).Conditionedmedia from SAG-stimulated GLI1+ cells, but not SAGalone, significantly increased the proliferation of serum-starvedmelanocytes (Fig. S3 e). These results indicate that, in addition topromoting mast cell survival, HH-stimulated choroidal GLI1+

cells may modulate the homeostasis of choroidal melanocytes.

HH signaling modulates the expression ofinflammation-related genes in choroidal GLI1+ cellsTo gain further insight into the crosstalk mechanisms andfunctional roles of the HH pathway in the choroid, we per-formed RNAseq analyses of cultured choroidal GLI1+ cells in theabsence or presence of the HH pathway agonist SAG (Fig. 5 a andData S1 j). Gene Ontology analyses of up- and down-regulatedgenes (at least twofold) using DAVID software (Huang et al.,2009) showed that the most significant transcriptome changesupon HH pathway activation corresponded to genes related toimmune response, including the categories cellular response toIFN-β, cellular response to IFN-γ, and innate immune response(Fig. 5 b). Gene set enrichment analysis (GSEA; Subramanianet al., 2005) confirmed that SAG treatment significantly re-duced the expression of genes related to immune response andcellular response to IFN-β and IFN-γ (Fig. 5 c). Moreover,Ingenuity Pathway Analysis (Qiagen) showed that HH pathwayactivation significantly inhibited transcriptional networksdownstream IFN-β, IFN-γ, several IFN regulatory factors, andproinflammatory receptors and transcriptional regulators suchas IL-6R, IL-1R1, STAT1, and NF-κB (Data S1 k). Conversely, HHactivation resulted in gene expression changes that mimickedtranscriptional events downstream of IL-10RA, the receptor forthe anti-inflammatory cytokine IL-10 (Data S1 k). Next, we

sought to assess the contribution of such crosstalk mechanismto the transcriptional changes induced in RPE/choroid tissueafter EC-specific Ihh deletion. Using our RNAseq data, we se-lected genes whose fold change expression in SAG-treatedGLI1+ cells (adj P < 0.05 versus control cells) was inverselycorrelated to the fold change observed in RPE/choroid fromIhhiΔEC/iΔEC mice (adj P < 0.05 versus Ihh+/+ mice; Fig. 5 d). Outof 158 genes significantly down-regulated in IhhiΔEC/iΔEC mice,29 were up-regulated in SAG-treated GLI1+ cells. As expected,these genes included HH downstream targets such as Gli1,Ptch1, Foxl1, and Hhip. On the other hand, RPE/choroid fromIhhiΔEC/iΔECmice presented increased expression of 181 genes, 15of which were down-regulated in SAG-treated GLI1+ cells. In-terestingly, genes with the highest differential expression ofthis list included the IFN-activated genes Ifi202b and Gm14446(Ifit1bl1), as well as F2rl1, which encodes the proinflammatoryreceptor PAR2 (Dery et al., 1998; Steinhoff et al., 2000). Col-lectively, our bioinformatics analyses strongly suggest that HHsignaling blunts the transcriptional changes of choroidal GLI1+

cells in response to proinflammatory factors. Thus, crosstalkbetween EC-secreted IHH and stromal GLI1+ cells may be keyfor the maintenance of choroidal immune homeostasis.

EC-specific Ihh deletion alters choroidal and retinalinflammatory responses and aggravates visual functionimpairment after tissue damageTo evaluate the impact of EC-specific Ihh deletion on choroidimmune cells, we first sought to identify the different cell typeswithin choroidal hematopoietic cells and to characterize theirmolecular identity.We reanalyzed our scRNAseq data using onlycells from cluster 11 (Fig. 1 b), and unsupervised cell clusteringresulted in five transcriptionally distinct subclusters (Fig. 6 aand Data S1 l). We assembled lists of specifically enrichedgenes (average normalized UMIs at least fivefold higher com-paredwith other hematopoietic subclusters; adj P < 0.05; Data S1m),which were used to assess the identity of each subcluster usingthe Immunological Genome Project website (Heng et al., 2008;http://www.immgen.org; Fig. S4 a). Such analysis allowed us toidentify them as macrophages (subcluster H2), T cells (H3),dendritic cells (DCs; H4), and mast cells (H5). The identity ofsubcluster H1 was not obvious, since some H1-specific geneswere found to be expressed in DCs, macrophages, monocytes,and neutrophils (Fig. S4 a). After close inspection of ourscRNAseq data, we classified H1 as Itgae (CD103)− Itgam(CD11b)+ classical DCs (cDCs) and H4 as CD103+ CD11b− cDCs(Merad et al., 2013; Fig. S4 b).

Next, we analyzed how EC-specific Ihh deletion modulateschoroidal immune cells in vivo by combining our scRNAseqcharacterization of choroidal cell types and our bulk RNAseqanalyses of RPE/choroid from IhhiΔEC/iΔEC mice. We selected

number of mast cells normalized by flatmount area and presented as relative units (Ihh+/+, n = 3; IhhiΔEC/iΔEC, n = 5, t test). *, P < 0.05. (f) In vitro assays to studythe survival of serum-starved RBL-2H3 cells (red) in the absence or presence of choroidal GLI1+ cells and the HH pathway agonist SAG. Dead cells are DAPI+

(green). (g)Quantification of RBL-2H3 cell death shown in panel f. The graph shows the percentage of DAPI+ RBL-2H3 cells. Black dots correspond to biologicalreplicates, and red bars represent average values (n = 3, ANOVA + Bonferroni test). Results are representative of two independent experiments. *, P < 0.05; ***,P < 0.001.



Dow



http://www.immgen.org

https://doi.org/10.1084/jem.20190730

genes specifically enriched in hematopoietic cells (cluster 11 inFig. 1 b; average normalized UMIs at least fivefold higher com-pared with other cell types; adj P < 0.05; Data S1 c) that weresignificantly reduced after Ihh deletion (adj P < 0.05; Data S1 h).To obtain more hits than those shown in Fig. 4 b (Cma1 andGpr34), we did not apply any FPKM or fold-change threshold tobulk RNAseq data. This analysis revealed that, in addition toCma1 and Gpr34, RPE/choroid expression of the hematopoietic-enriched genes Cd163, Pf4 (CXCL4), Mrc1 (CD206), and Cd72 wassignificantly reduced after EC-specific Ihh deletion (Fig. 6 b).With the exception of Cma1 (specifically expressed in mast cells,as expected), the rest of hematopoietic-enriched genes that weredown-regulated in RPE/choroid tissue from IhhiΔEC/iΔEC miceweremarkedly enriched in themacrophage population (Fig. 6 c).Interestingly, Cd163 and Mrc1 (CD206) are well-establishedmarkers of alternatively activated M2 macrophages, which aregenerally involved in inflammation resolution and tissue re-modeling (Biswas and Mantovani, 2010; Murray, 2017). Weperformed immunofluorescence assays and quantitative imaginganalyses onRPE/choroid flatmounts fromalbino Ihh+/+ and IhhiΔEC/iΔEC

mice using antibodies against the murine pan-macrophage

marker F4/80 and the M2 marker CD206. EC-specific Ihh dele-tion did not alter the percentage of F4/80+ area, suggesting thepresence of a similar number of macrophages than in controlanimals (Fig. 6, d and e). However, IhhiΔEC/iΔEC mice showed re-duced CD206 expression in F4/80+ cells (Fig. 6, d and f). Insummary, our observations suggest that (1) the main choroidalhematopoietic cell types affected by EC-specific Ihh deletion aremast cells andmacrophages and (2) EC-specific Ihh deletion skewsmacrophage polarization away from the anti-inflammatory M2phenotype.

In light of these results, we reasoned that inducing RPE/choroid injury in IhhiΔEC/iΔEC mice might result in a longer ormore pronounced inflammatory response than in control ani-mals. To test this hypothesis, we performed experiments basedon NaIO3 administration, a well-established model of retinalinjury that induces selective RPE death in a dose-dependentmanner and subsequent photoreceptor loss and vision impair-ment (Kannan and Hinton, 2014). We injected i.v. a single, lowdose (15 mg kg−1) of NaIO3 into Ihh+/+ and IhhiΔEC/iΔEC mice, and3 d later, RNA was extracted from RPE/choroid and neuralretina to assess the expression of proinflammatory markers by

Figure 5. HH signaling modulates the ex-pression of inflammation-related genes inchoroidal GLI1+ cells. (a) Experiment outline.GLI1+ cells (GFP+) were isolated from RPE/cho-roid tissue of Gli1GFP/+ mice by FACS. Culturedcells were exposed to the HH pathway agonistSAG (100 nM) or vehicle (DMSO) for 24 h, andgene expression was analyzed by RNAseq.(b) Biological process (GOTERM_BP_DIRECT)Gene Ontology analysis of control- and SAG-treated GLI1+ cells using DAVID software, con-sidering genes with log2 fold change ≤−1 and≥1 (n = 3, Benjamini–Hochberg corrected adj P <0.05). Significantly enriched gene sets are shown(Benjamini-corrected adj P < 0.05). (c) Pre-ranked GSEA shows that SAG treatment re-duces the expression of immune-related genes.FDR, false discovery rate q-value; NES, normal-ized enrichment score; Nom. P, nominal P value.(d) Genes with inversely correlated differentialexpression between HH activation in vitro (GLI1+

cells after SAG treatment) and inhibition in vivo(RPE/choroid tissue after EC-specific Ihh deletion).Numbers in Venn diagrams indicate the number ofgenes significantly up- or down-regulated in eachcondition (Benjamini–Hochberg corrected adj P <0.05). Gray intersections indicate the number ofoverlapping genes for each comparison, which arelisted on the graph on the right. Green bars rep-resent genes up- or down-regulated in SAG-treated GLI1+ cells compared with control cells,and red bars represent genes up- or down-regulatedin RPE/choroid tissue from IhhiΔEC/iΔEC mice com-pared with Ihh+/+ mice.



Dow



https://doi.org/10.1084/jem.20190730

real-time PCR.We observed a trend toward increased expressionof all genes tested in both RPE/choroid and neural retina fromIhhiΔEC/iΔECmice, reaching statistical significance for Ccl2 and Tnfin both tissues and Il6 and Cxcl10 in neural retina (Fig. 6 g). Totest the potential functional consequences of such augmented

proinflammatory profile, we performed optomotor tracking as-says to assess visual function in Ihh+/+ and IhhiΔEC/iΔEC mice be-fore and 3 d after NaIO3 administration. In the absence of NaIO3

treatment, EC-specific Ihh deletion did not compromise visualfunction (Fig. 6 h). However, NaIO3 treatment induced a

Figure 6. EC-specific Ihh deletion alters choroidal and retinal inflammatory responses and aggravates visual function impairment after tissuedamage. (a) tSNE graph showing five transcriptionally distinct subclusters (H1–H5) within choroidal hematopoietic cells, as determined by scRNAseq.(b) RNAseq data showing genes significantly enriched in choroidal hematopoietic cells (as determined by our scRNAseq results) and down-regulated in RPE/choroid from IhhiΔEC/iΔEC mice (blue dots) compared with Ihh+/+ mice (black dots). Individual dots correspond to different animals, and red bars representaverage values. Relative gene expression is presented as the percentage of Ihh+/+ mice (n = 3, Benjamini–Hochberg corrected adj P value). *, P < 0.05; **, P <0.01; ***, P < 0.001. (c) Relative expression levels of the genes shown in panel b among choroidal hematopoietic subclusters calculated using the averagenormalized UMIs in each subcluster and represented as the percentage of the subcluster with maximum expression. White, 0%; red, 100%. (d) Immuno-fluorescence analysis of F4/80 and CD206 expression in RPE/choroid flatmounts from albino Ihh+/+ and IhhiΔEC/iΔEC mice. Scale bar, 1,000 µm. Panels on theright show zoomed-in regions (scale bar, 125 µm). (e) Quantification of F4/80+ area shown in panel d. Results are normalized by flatmount area and presentedas relative units (Ihh+/+, n = 3; IhhiΔEC/iΔEC, n = 6, t test). (f) Quantification of CD206 intensity within F4/80+ area shown in panel d. Results are presented asrelative units (Ihh+/+, n = 3; IhhiΔEC/iΔEC, n = 6, t test). *, P < 0.05. (g) Real-time PCR assays showing relative expression levels of pro-inflammatory genes in RPE/choroid (left) and neural retina (right) from Ihh+/+ and IhhiΔEC/iΔECmice 3 d after i.v. administration of 15mg kg−1 NaIO3. Results are presented as in panel a (n = 4,t test). *, P < 0.05; **, P < 0.01; ***, P < 0.001. (h) Optomotor response analyses of Ihh+/+ and IhhiΔEC/iΔEC mice before (day 0) and after (day 3) NaIO3 ad-ministration (day 0, n = 7; day 3, n = 10; ANOVA + Bonferroni test). *, P < 0.05; ***, P < 0.001. c/d, cycles per degree.



Dow



https://doi.org/10.1084/jem.20190730

significantly impaired spatial visual function of IhhiΔEC/iΔEC micecompared with Ihh+/+ mice. These results demonstrate that in acontext of tissue damage, EC-specific Ihh deletion results in anexacerbated inflammatory response in the RPE/choroid andneural retina that correlates with a more severe visual functionimpairment.

DiscussionHerein, we report the transcriptome of mouse RPE/choroidtissue at single cell resolution and a study of HH signaling in theadult choroid. Taken together, our results support a model inwhich choroidal EC-expressed Ihh and HH-respondingMSC-likecells are key for several aspects of choroidal and retinal immunehomeostasis, including survival of choroidal mast cells and in-flammatory response after tissue damage. Our findings consti-tute a new angle for the study of inflammation-related oculardisorders such as AMD.

A main contribution of our work is the identification andmolecular characterization of transcriptionally distinct subtypesof choroidal ECs. As increasing evidence indicates that micro-vascular ECs are tissue specific and secrete specialized sets ofangiocrine factors that regulate organ homeostasis and regen-eration (Rafii et al., 2016), it was intriguing to find that chori-ocapillaris ECs, located next to RPE cells, exhibit a highly specifictranscriptional profile, including a marked enrichment in Ihhexpression. Previous work in the developing mouse eye hasshown that Ihh is expressed by a subset of choroidal cells adja-cent to the RPE presumed to be ECs based on the expression ofcollagen IV (Dakubo et al., 2003, 2008; Wallace and Raff, 1999).However, whether Ihh expression is maintained in the adultchoroid remained unknown so far. Our scRNAseq and bulkRNAseq analyses unambiguously demonstrate that Ihh is ex-pressed at high levels in adult mouse choriocapillaris ECs rela-tive to other ECs in the choroid, retina, and extraocular tissues.The abovementioned study showed that Ihh knockout mice(which die during embryonic stages or at birth; St-Jacques et al.,1999) display defects in RPE, sclera, and neural retina prenataldevelopment (Dakubo et al., 2008); now, we provide evidencethat Ihh expression also plays an important role in adult choroidhomeostasis.

Experiments using reporter Gli1GFP/+ mice revealed that themain target of choroidal HH signaling is a population of peri-vascular cells that molecular, cellular, and functional assaysidentified as GLI1+ MSC-like stromal cells, in agreement withprevious findings in other organs (Kramann et al., 2015; Zhaoet al., 2014). Importantly, response of these cells to HH signalingseems to regulate the homeostasis of choroidal mast cells andmelanocytes. It remains to be determined the role of this sig-naling circuit in the human choroid and its potential disease-associated alterations; nevertheless, the well-establishedimmunomodulatory ability of MSCs (Uccelli et al., 2008) andthe involvement of inflammation in several choroid-related eyepathologies such as AMD, diabetic retinopathy, and uveitissuggest that the choroid HH pathway we have uncovered mayhelp in the understanding and treatment of such diseases. In-terestingly, transcriptional analysis of RPE/choroid tissue from

AMD donors and age-matched controls showed that expressionof the proinflammatory genes CCL2 and CXCL10, which is up-regulated upon tissue damage after EC-specific Ihh deletion(Fig. 6 g), is elevated in all major clinical AMD phenotypes(Newman et al., 2012). The role of HH signaling in the controland resolution of inflammation has been described in severalpathological contexts, especially in the intestine, at the level ofboth HH ligands and target cells. In humans, a GLI1 non-synonymous polymorphism is strongly associated with ulcera-tive colitis, where areas of colonic inflammation displayreduced expression of the HH target genes GLI1, PTCH, andHHIP (Lees et al., 2008). In mice, specific deletion of Ihh fromenteric cells induces the expression of inflammation-relatedgenes, leukocyte infiltration of the crypt area, and the devel-opment of intestinal fibrosis (van Dop et al., 2010; Westendorpet al., 2018). In agreement with our findings in the choroid, themain target of intestinal HH signaling is a population offibroblast-like stromal cells (Westendorp et al., 2018). More-over, inhibition of intestinal HH signaling by overexpression ofthe HH inhibitor HHIP induces spontaneous inflammation, di-arrhea, weight loss, and death (Zacharias et al., 2010). Using amouse model of chemically induced colitis, it was shown thatinhibition or activation of the HH pathway intensifies orameliorates colon inflammation, respectively (Lee et al., 2016).Interestingly, this study showed that HH activation induces Il10expression by stromal cells and results in increased numbers ofintestinal CD4+FOXP3+ regulatory T cells. Given the connectionbetween regulatory T cells and mast cells (see below), it wouldbe interesting to study whether in the intestine, similarly toour observations in the choroid, HH signaling regulates thepresence of mast cells. Importantly, intestinal mesenchymeresponds to HH signaling in vitro by down-regulating proin-flammatory genes (Zacharias et al., 2010), in line with ourfindings using cultured choroidal GLI1+ cells. Furthermore, theHH pathway acts as an endogenous anti-inflammatory systemat the level of the blood–brain barrier (Alvarez et al., 2011). Wenow add the choroid to the list of adult tissues where the in-flammatory response is modulated by HH signaling.

Additional in vivo and in vitro experiments indicated thatHH-stimulated MSC-like cells are key for choroidal mast cellhomeostasis and survival. Although mast cells are present inthe choroid of humans and most vertebrates (May, 1999;McMenamin, 1997; McMenamin and Polla, 2013), their role inthe eye in health and disease remains unknown. AMD is asso-ciated with an increased number of total and degranulated mastcells in the choroid (Bhutto et al., 2016; McLeod et al., 2017).However, the interpretation of these observations is unclear, asmast cells not only participate in proinflammatory responses butalso can limit inflammation or facilitate its resolution (Caughey,2011; Galli et al., 2008; Tsai et al., 2011). Mast cells are key for theimmunosuppressive effect of UVB radiation (Hart et al., 1998)and the induction of tolerance to skin allografts mediated byCD4+FOXP3+ regulatory T cells (Lu et al., 2006). Interestingly,genetic mast cell depletion results in increased numbers of in-filtrating immune cells in the eye after endotoxin-induceduveitis, suggesting that choroidal mast cells may play a protec-tive, anti-inflammatory role in this organ (Smith et al., 1998).



Dow



https://doi.org/10.1084/jem.20190730

Indeed, our scRNAseq data show that mast cells are the onlychoroidal cell type with detectable expression of Il4 (Data S1, band m). Given the well-established role of IL-4 in promotingmacrophage polarization toward the anti-inflammatory M2phenotype (Biswas and Mantovani, 2010; Murray, 2017), it ispossible that mast cell loss contributes to the reduction in M2markers we observed in choroidal macrophages from IhhiΔEC/iΔEC

mice (Fig. 6, c and d–f). Hence, our results suggest that im-pairment of HH signaling may impact the pathogenesis of in-flammatory choroid diseases through a decrease in the populationof choroidal mast cells.

Because RPE and photoreceptor death is a hallmark of AMD,significant progress has been made to generate stem cell–derived RPE and photoreceptors for transplantation purposes(Nazari et al., 2015). However, as choriocapillaris loss is an earlyevent during the development of AMD (Bhutto and Lutty, 2012;Whitmore et al., 2015), and because oxygen and nutrient supplyof the outer retina depends on choroidal perfusion, it is likelythat long-lasting engraftment and proper function of trans-planted retinal cells could be significantly improved by co-transplantation of choriocapillaris ECs. However, the generationof such specialized ECs is challenging, in part due to the lack ofinformation about their molecular identity. To date, only onelaboratory has reported the generation of stem cell–derivedchoroid EC-like cells, and the assessment of their resemblance toauthentic choriocapillaris ECs was limited to the expression of afew markers (Songstad et al., 2015, 2017). Although our char-acterization of choroidal EC subtypes by scRNAseq needs furthervalidation in human tissue, it demonstrates marked gene ex-pression differences between the choriocapillaris and otherchoroidal endothelial layers. We believe such transcriptomeheterogeneity should be taken into account when generatingstem cell–derived choriocapillaris ECs suitable for cell replace-ment strategies. Moreover, cell surface proteins encoded bychoroidal EC signature genes (Fig. 2 c) constitute potential tar-gets for site-directed delivery of functionalized nanoparticles orviral vectors after systemic administration, which could open newtherapeutic approaches to treat choroidal and retinal pathologies.

Materials and methodsscRNAseq of mouse RPE/choroid tissueExperiments were performed with 90-d-old mice of two dif-ferent strains, C57BL/6J and B6129PF1/J. To isolate viable, singlecells from RPE/choroid tissue, we euthanized the mice andwashed the blood via intracardial perfusion with PBS buffer, andthen enucleated the eyes for further processing. Each eye wascleaned from extraocular tissue and a circumpherencial incisionwas performed right below the ora serrata to remove the ante-rior segment, including cornea, lens, iris and ciliary body. Theneural retina was detached from the optic nerve and RPE/cho-roid tissue was gently scraped from the sclera. For each strain,experiments included one eye from a male and another from afemale, which were mixed as a strategy to estimate the per-centage of contaminating doublets in our preparation. RPE/choroid tissue was incubated with Collagenase A (6.25 mg ml−1),Dispase II (6.25 mg ml−1), and DNase (62.5 µg ml−1; Roche) at

37°C for 15 min as previously described (Benedicto et al., 2017)and then in 0.25% trypsin-EDTA (Gibco) at 37°C for 5 min tocreate a single cell suspension. Both solutions also includedHoechst (10 µM; Thermo Fisher Scientific) to label nuclei duringdigestion. The cell suspension was pelleted, resuspended insorting buffer (PBS plus 5% FBS and 2 mM EDTA), filteredthrough a 70-µm cell strainer, stained with TO-PRO-3 (1 µg ml−1;Thermo Fisher Scientific), and purified by FACSwith gates set toinclude nucleated particles (Hoechst+) and exclude dead cells(TO-PRO-3+).

Library preparation, sequencing, and raw data after pro-cessing were performed at the Weill Cornell Medicine Epi-genomics Core. scRNAseq libraries were prepared according to10x Genomics specifications (Single Cell 39 Reagent Kits v2 UserGuide PN-120236; 10x Genomics). Cellular suspensions at aconcentration between 250 and 500 cells/µl were loaded ontothe 10x Genomics Chromium platform to generate barcodedsingle-cell gel bead-in-emulsion (GEMs), targeting ∼5,000 sin-gle cells per sample. GEM-reverse transcription (53°C for45 min, 85°C for 5 min) was performed in a C1000 Touchthermal cycler with 96-Deep Well Reaction Module (Bio-Rad).Next, GEMs were broken and the single-strand cDNA wascleaned upwith DynaBeadsMyOne Silane Beads (Thermo FisherScientific). The cDNAwas amplified (98°C for 3 min; 98°C for 15 s,67°C for 20 s; and 72°C for 1 min × 12 cycles) using the C1000Touch Thermal cycler with 96-Deep Well Reaction Module.Quality of the cDNA was assessed using an Agilent Bioanalyzer2100, obtaining a product of ∼1,170 bp. Amplified cDNA was en-zymatically fragmented, end repaired, A-tailed, subjected to adouble-sided size selection with SPRIselect beads (BeckmanCoulter), and ligated to adaptors provided in the kit. A uniquesample index for each library was introduced through PCR am-plification using the indexes provided in the kit (98°C for 45 s;98°C for 20 s, 54°C for 30 s, and 72°C for 20 s × 14 cycles; and 72°Cfor 1 min). Indexed libraries were subjected to a second double-sided size selection, and libraries were then quantified usingQubit fluorometric quantification (Thermo Fisher Scientific).The quality was assessed on an Agilent Bioanalyzer 2100, ob-taining an average library size of 455 bp.

Libraries were diluted to 2 nM and clustered on an IlluminaHiSeq2500 high-output mode at 10 pM on a pair-end read flowcell and sequenced for 26 cycles on R1 (10x barcode and UMIs),followed by 8 cycles of I7 Index (sample Index), and 98 bases onR2 (transcript), with a coverage ∼145 million reads per sample.Primary processing of sequencing images was done using Illu-mina’s Real-Time Analysis software. 10x Genomics Cell RangerSingle Cell Software suite v2.1.0 was used to perform sampledemultiplexing, alignment (mm10), filtering, UMI counting,single-cell 39 end gene counting, and quality control using themanufacturer parameters. 3,996 (C57BL/6J mice) and 3,727(B6129PF1/J mice) single cells were sequenced with ∼34,000mean reads (∼75% sequencing saturation) and a median of∼1,500 detected genes per cell.

Unsupervised cell clustering analysis was performed usingthe Seurat 2.1 R package (Butler et al., 2018). Cells with <1,700UMIs, <600 genes, or >10% mitochondrial genes were excludedfrom the analysis, as well as genes detected in less than three



Dow



https://doi.org/10.1084/jem.20190730

cells. Gene expression raw counts were normalized following aglobal-scaling normalization method with a scale factor of10,000 and a log transformation using the Seurat NormalizeDatafunction. The top 2,000 highly variable genes from C57BL/6Jand B6129PF1/J datasets were selected, followed by a canonicalcorrelation analysis (CCA) to identify common sources of vari-ation between the two datasets and minimize batch effect. Thefirst 11 aligned CCA results were used to generate two-dimensional tSNE (RunTSNE in Seurat, perplexity = 30; Vander Maaten, 2008) and unsupervised cell clustering by a sharednearest neighbor (FindClusters in Seurat, k.param = 30 andresolution = 0.8). A list of conserved cell-type–specific geneswere generated by FindConservedMarkers (test.use = “bimod”,logfc.threshold = 0.5, min.pct = 0.25) function in Seurat, iden-tifying ≥65 conserved cell-type–specific genes for each cluster inboth mouse strains (adj P < 0.05). Cells contained in cluster 11(hematopoietic cells) were further subjected to a second round ofunsupervised analysis following the same approach using theSeurat FindClusters with ∼1.3 resolution. The modified Seuratfunction FindAllMarkers was used to identify differentially ex-pressed genes among subclusters within endothelial and hema-topoietic cells. P values were calculated using likelihood-ratiotest for single-cell gene expression (McDavid et al., 2013) andadjusted by the Benjamini–Hochberg method. More detailedarguments are available on Github (see Code Availability section).

Estimation of doublets was calculated based on the exclusiveexpression of Xist and Ddx3y in female and male cells, respec-tively. Expression of both genes associated to the same cellularbarcode indicates the presence of a doublet. We made the fol-lowing assumptions: (1) because we pooled the same amount oftissue from each sex before RPE/choroid digestion, 50% cells aremale and 50% cells are female; (2) female–female and male–maledoublet rates are the same as the female–male doublet rate; and(3) female–male doublets account for 50% of total doublets, sincethe other 50% are female–female and male–male doublets. Wecalculated Xist capture rate using the following equations:

Xist+ cells � [(Xist+Ddx3y− cells) + (Xist+Ddx3y+ cells)]Xist+ capture rate � (Xist+ cells

�all cells) × 2.

Note that the capture rate result was multiplied by 2 becauseonly 50% of all cells are expected to be female. Similar equationswere used to calculate Ddx3y capture rate:

Ddx3y+ cells � [(Xist−Ddx3y+ cells) + (Xist+Ddx3y+ cells)]Ddx3y+ capture rate � (Ddx3y+cells

�all cells) × 2.

Female–male doublet rate was calculated using the followingequation:

(All cells) × (Xist+ capture rate) × (Ddx3y+ capture rate)× (female −male doublet rate) � Xist+Ddx3y+ cells.

Because female–male doublet rate accounts for 50% of totaldoublets, the corrected doublet rate results from multiplying by2 the female–male doublet rate. Calculations are shown on Fig. S1 b.

Code availabilityCustom code has been deposited at GitHub (https://github.com/nyuhuyang/scRNAseq-MouseEyes).

Immunofluorescence assays and quantitative image analysisFor cryosection staining, mouse eyes were enucleated andtransferred to 4% paraformaldehyde in PBS. A hole was made inthe cornea with a 22G needle, and cornea and lens were re-moved. Eyecups were incubated in 4% paraformaldehyde for 2 hat room temperature. Eyes were washed in PBS, incubated in15% sucrose in PBS for 30 min, and kept overnight at 4°C in 30%sucrose in PBS. Eyecups were infiltrated in Tissue-Tek optimalcutting temperature compound (Sakura) and frozen in the samesolution at −80°C. Sections 5–10 µm thick were obtained with acryostat (Hacker) and mounted on Superfrost Plus Gold Micro-scope Slides (Thermo Fisher Scientific). For immunostaining,eye sections were washed twice for 5 min in PBS, incubated for1 h at room temperature in blocking buffer (PBS supplementedwith 3% bovine serum albumin and 0.3% Triton X-100; Sigma)and incubated overnight at 4°C in blocking buffer plus the fol-lowing primary antibodies: rat anti-mouse KDR (clone Avas12,1:100, catalog number 136401; BioLegend), mouse anti-claudin 5(clone 4C3C2, 1:500, catalog number 352588; Invitrogen), mouseanti-RPE65 (clone 401.8B11.3D9, 1:100, catalog number NB100-355; Novus Biologicals), rat anti-mouse CD140b (PDGFRβ; 1:200,catalog number 14–1402-81; Invitrogen), and/or rabbit anti-GFP(1:1,000, catalog number A11122; Invitrogen). After three washesin PBS for 10 min each, sections were incubated for 1 h at roomtemperature with Alexa Fluor 647 goat anti-rat and/or AlexaFluor 488 goat anti-rabbit or anti-mouse antibodies (1:500; LifeTechnologies) in blocking buffer, washed three times in PBS for10 min each, and mounted with Vectashield (Vector Laborato-ries). Intravital labeling of ECs was performed by retroorbitali.v. delivery of 20 µg Alexa Fluor 647–conjugated rat anti-mouseVE-cadherin (clone BV13, catalog number 138006; BioLegend) inmice previously anesthetized using isoflurane. After 10 min,mice were perfused with 4% paraformaldehyde and euthanized,and eyeballs were enucleated and processed for immunofluo-rescence as described above. To stain mast cells and macro-phages in RPE/choroid flatmounts from albino mice, animalswere dark-adapted overnight for better detachment of retinafrom the RPE and euthanized the following day. Eyes wereenucleated and fixed in 4% paraformaldehyde in PBS for 2 h atroom temperature, and cornea and lens were removed duringfixation. Eyecups were washed twice in PBS and retina wascarefully removed. RPE/choroid tissue was processed andstained as described above using Rhodamine Avidin D, TMRITC(1:250, catalog number A-2002; Vector Laboratories) for mastcell detection (Tharp et al., 1985; Veerappan et al., 2008), or ratanti-mouse F4/80 (1:500, catalog number ab6640; Abcam) andrabbit anti-mouse CD206 (1:500, catalog number ab64693; Ab-cam) plus Alexa Fluor 647 goat anti-rat and Alexa Fluor 568 goatanti-rabbit for macrophage staining. Images were collected witha Zeiss Axio Observer spinning disk confocal microscopeequipped with a Yokogawa scanner unit, Hamamatsu Evolveelectron-multiplying charge-coupled device cameras (Photo-metrics), and either a 63× (Fig. 3 c) or 20× (for flatmounts) ob-jective. Images from Fig. 1 e were acquired using a Zeiss AxioObserver LSM 880 confocal microscope with a Plan Apochromat40×/numerical aperture 0.95 objective. Mast cell quantificationwas performed using the open-source software Ilastik, which



Dow



https://github.com/nyuhuyang/scRNAseq-MouseEyes

https://github.com/nyuhuyang/scRNAseq-MouseEyes

https://doi.org/10.1084/jem.20190730

implements a random forest supervised machine learning clas-sifier (Berg et al., 2019). Briefly, an initial training was per-formed using a control flatmount to separate foreground objects(mast cells) from the background and then applied to the re-maining samples. The segmented binary images were exportedand further processed in ImageJ using fill holes and watershedfunctions and then quantified using the analyze particles func-tion. Mast cell numbers were normalized to the total area foreach flatmount. Macrophage quantification was performed withImageJ software. After background subtraction (20 pixels), F4/80 signal was thresholded (default parameters) and used tocreate a selection, which was normalized by total flatmount areato calculate F4/80+ area. The same F4/80+ selection was appliedto the background-subtracted CD206 image. CD206 mean in-tensity was measured within the F4/80+ selected area withoutapplying any threshold.

Isolation and RNAseq of tissue-specific mouse ECsMouse tissue-specific ECs were isolated from 30-d-oldB6129PF1/J mice by intravital staining of VE-cadherin followedby tissue digestion and cell sorting as previously described(Benedicto et al., 2017). Cells were lysed immediately aftersorting and processed for RNAseq as previously described(Benedicto et al., 2017). RNAseq was performed from three in-dependent isolations, with 7 animals (14 eyes) per isolation forRPE/choroid and neural retina, and one mouse per isolation forliver, lung, and heart. cDNA libraries were prepared with theTruSeq RNA Sample Preparation Kit (Illumina) and sequencedon an Illumina HiSeq2000 platform. Upon quality control usingFastQC, raw reads were aligned to the mouse genome (mm10),downloaded via the UCSC genome browser (http://hgdownload-test.cse.ucsc.edu/goldenPath/mm10/bigZips/) using STAR_2.4.0f1 (Dobin et al., 2013) and SAMTOOLS v0.1.19 (Li et al., 2009)for sorting and indexing reads. Cufflinks (2.0.2; Trapnell et al.,2012) was used to get the expression values (FPKM). Log-transformed FPKM profiles were clustered using hierarchicalclustering (hclust function in R language) with average linkageand one minus Spearman correlation as distance. To determinethe genes differentially expressed between choroidal ECs andECs from neural retina, lung, liver, and heart, statistical signif-icance was calculated using DEseq2 bioconductor package in R,which uses read counts (htseq-count; Love et al., 2014). With adetection threshold set at ≥1 FPKM, genes expressed in choroidalECs at levels at least fivefold higher compared with ECs fromneural retina, lung, liver, and heart (Benjamini–Hochberg cor-rected adj P < 0.05) were considered choroidal EC signature genes.

Gene expression analyses of mouse ECs and RPE/choroidtissue by real-time PCRFor tissue-specific native ECs, RNA was extracted immediatelyafter sorting using TRI Reagent (Molecular Research Center) andthe RNeasyMini Kit (Qiagen) as previously described (Benedictoet al., 2017). For mouse RPE/choroid tissue, mice were eutha-nized, and eye globes were enucleated followed by careful ex-cision of any extraocular tissue that remained attached to thesclera. The anterior segment was discarded, and after removalof the neural retina, RPE/choroid tissue was mechanically

dissected from the sclera using a scalpel and immediately usedfor RNA extraction with the RNeasy Mini Kit (Qiagen). cDNAwas prepared with the High Capacity cDNA Reverse Transcrip-tion Kit (Life Technologies), and real-time PCRwas performed ina StepOnePlus Real-Time PCR System (Life Technologies) usingSYBR Select Master Mix (Life Technologies) and the followingmouse primer pairs: Ihh (NM_010544, 59-ACGTGCATTGCTCTGTCAAG-39 and 59-GTCTCCTGGCTTTACAGCTG-39),Gli1 (NM_010296, 59-CCCAGCTCGCTCCGCAAACA-39 and 59-CTGCTGCGGCATGGCACTCT-39), Cma1 (NM_010780, 59-CAAGCCTGCAAACACTTCAC-39 and 59-CATAGGATGCAATGCCTTGG-39), Mcpt4 (NM_010779, 59-ATCTGGAGATCACCACTGAG-39 and 59-TCACATCATGAGCTCCAAGG-39), Tpsb2 (NM_010781, 59-CCATTGTCCATGATGGCATG-39 and 59-TGTAGATGCCAGGCTTGTTG-39), Fabp4 (NM_024406, 59-GTGATGCCTTTGTGGGAACC-3 and 59-TCATGTTGGGCTTGGCCATG-39), Spp1 (NM_001204233, 59-TGCCTGACCCATCTCAGAAG-3 and 59-TCGTCGTCCATGTGGTCATG-39), Il6 (NM_031168, 59-GAACAACGATGATGCACTTGC-39 and 59-TCCAGGTAGCTATGGTACTC-39), Ccl2 (NM_011333, 59-CATTCACCAGCAAGATGATCC-3 and 59-TGTATGTCTGGACCCATTCC-39), Nos2 (NM_001313921, 59-TGCCTCATGCCATTGAGTTC-3 and 59-AGATGAGCTCATCCAGAGTG-39), Tnf (NM_013693, 59-CTCTTCTCATTCCTGCTTGTG-3 and 59-TCTGGGCCATAGAACTGATG-39), and Cxcl10(NM_021274, 59-AGTGAGAATGAGGGCCATAG-3 and 59-TCCGGATTCAGACATCTCTG-39). Gapdh (NM_008084, 59-AAATGGTGAAGGTCGGTGTG-39 and 59-GAGTGGAGTCATACTGGAAC-39) wasused as loading control, and relative expression values werecalculated by the 2−ΔΔCt method.

Assessment of IHH expression in human tissuesHuman retina and combined RPE/choroid punches were se-lected from a research collection at the University of Iowa In-stitute for Vision Research. This collection contains tissuesobtained from the Iowa Lions Eye Bank (Iowa City, IA) after fullconsent of the donors’ families. All experiments conformed tothe Declaration of Helsinki. The eyes used in this study had noknown ocular history of AMD. Eyes were dissected by making acircumferential incision behind the limbus and removing theanterior segment consisting of cornea, iris, lens, and ciliarybody. The remaining posterior pole was cut into a clover-leafshape by making four equally spaced incisions through thesclera, choroid, RPE, and retina. An 8-mm sterile disposabletrephine punch, centered on the fovea centralis, was collectedfrom the macular region (easily identified by the presence offoveal pigment, proximity to the optic nerve head, and locationbetween the superior and inferior arcades in the temporal ret-ina). Neural retina and RPE/choroid samples were flash frozenseparately in liquid nitrogen, and all procedures were completedwithin 5.5 h after death. Punches were stored at −80°C untilused for RNA isolation. Samples were removed from the −80°Cfreezer and placed immediately on dry ice, such that theythawed in the presence of RNA extraction buffer (as even briefthawing is extremely destructive to RNA yield). Total RNA fromRPE/choroid and neural retina was extracted using the RNeasyMini Kit (Qiagen). Alternatively, total RNA from human lung,liver, and heart was obtained from Takara. Reverse transcrip-tion was performed in a 50-µl reaction using a reverse



Dow



http://hgdownload-test.cse.ucsc.edu/goldenPath/mm10/bigZips/

http://hgdownload-test.cse.ucsc.edu/goldenPath/mm10/bigZips/

https://doi.org/10.1084/jem.20190730

transcription kit (MultiScribe; Thermo Fisher Scientific). IHHexpression was determined by real-time PCR (NM_002181, 59-ATGACCCAGCGCTGCAAG-39 and 59-CAAAGCCGGCCTCCACTG-39)as described previously (Zeng et al., 2012) using RPL19 expressionfor normalization (NM_000981, 59-ATGCCAACTCCCGTCAGC-39and 59-ACCCTTCCGCTTACCTATGC-39) and calculating relativeexpression values by the 2−ΔΔCt method. All samples were ana-lyzed in technical triplicates.

Characterization of GLI1+ cells by flow cytometryGli1GFP/+ reporter mice in which GFP expression is controlledby the endogenous Gli1 promoter are described elsewhere(Brownell et al., 2011). Because these animals were generated inthe Swiss strain (SWR/J), they present retinal degeneration dueto the homozygous rd1mutation in the Pde6b gene (Chang et al.,2002). To avoid having confounding results due to retinal de-generation, we outcrossed these animals with wild-type BALB/cmice and intercrossed the offspring for at least two generationsto remove the rd1 phenotype before further characterization.RPE/choroid tissue from Gli1GFP/+ mice was mechanically dis-sected from the sclera as described above and digested in colla-genase A, dispase II and DNase as previously reported (Benedictoet al., 2017). Lungswere digested in the same conditions and usedto compare the percentage of GFP+ cells. For staining, cells fromtwo eyes were pelleted at 400 g for 5 min and resuspended inFACS buffer (PBS, 5% FBS, and 2 mM EDTA) including a CD16/CD32 antibody to block Fc receptors (clone 93, 1:200, catalognumber 101302, BioLegend). Cells were filtered through a 40-µmcell strainer and incubated for 30 min at room temperature withthe following fluorochrome-conjugated antibodies (BioLegend):CD45-APC/Cy7 (1:200, catalog number 103115), CD31-BV605 (1:200, catalog number 102427), CD29-PE/Cy7 (1:800, catalognumber 102221), CD105-PerCP/Cy5.5 (1:800, catalog number120415), PDGFRβ-PE (1:200, catalog number 136005), and SCA1-BV711 (1:200, catalog number 108131). Cells were washed threetimes in FACS buffer and resuspended in FACS buffer with 1 µg/ml DAPI to stain dead cells and exclude them from the analysis.Compensation controls were prepared using antibody-stainedUltraComp eBeads (Thermo Fisher Scientific). Cells and beadswere analyzed using a BD Fortessa flow cytometer and resultswere analyzed with FlowJo software.

Generation of EC-specific Ihh knockout miceGeneration of EC-specific Ihh inducible knockout was achievedby breeding IhhloxP/loxP mice (Jackson Laboratories) with Cdh5-CreERT2 transgenic mice (Wang et al., 2010) to establish theCdh5-CreERT2 IhhloxP/loxP line (IhhiΔEC/iΔEC). To induce Cre activity,animals were i.p. treatedwith tamoxifen at a dose of 250mg kg−1

in sunflower oil for 6 d (interrupted for 3 d after the third dose).After 8–12 wk of tamoxifen treatment, RPE/choroid tissue wasprocessed for RNAseq, real-time PCR, or imaging assays. AlbinoIhh+/+ and IhhiΔEC/iΔEC were generated by crossing pigmentedIhhiΔEC/iΔEC mice with albino Gli1+/+ mice already free of rd1mutation (see above). Heterozygous agouti (brown color) Ihh+/iΔEC

mice were obtained and crossed to generate albino homozygousIhhiΔEC/iΔEC mice. Second-generation animals were used for eitherexperiments or further breeding. In all experiments, tamoxifen-

injected littermate Cdh5-CreERT2 Ihh+/+mice were used as controls,and both male and female animals were used.

Culture, differentiation, and RNAseq analysis of choroidalGLI1+ cellsSorting experiments to isolate Gli1+ cells were performed at theWeill Cornell Medicine Flow Cytometry Core. RPE/choroid tis-sue from six Gli1GFP/+ mice (female, 72 d old) was dissected anddigested in collagenase A, dispase II, and DNase as describedabove. GFP+ live cells (DAPI−) were isolated with a BD Influx cellsorter into a well of a p24 plate containing 1 ml GLI1+ mediumcontaining MEM α modification (Sigma) plus penicillin/strep-tomycin 50× solution (Corning), Glutamax 100× solution, and20% FBS (Life Technologies). Cells were expanded up to threepassages and frozen until further use.

For differentiation experiments, cells were cultured inStemXVivo Osteogenic/Adipogenic Base Media supplementedwith either 100× adipogenic or 20× osteogenic supplement (R&DSystems) following themanufacturer’s instructions. After 21 d ofdifferentiation, cells were processed for real-time PCR to assessthe expression of adipogenic (Fabp4) and osteogenic (Spp1)markers (see above). Adipogenesis and osteogenesis were alsomonitored by staining with Oil Red O and Alizarin Red S(Sigma), respectively. At day 21, cells were washed in PBS andfixed in 4% paraformaldehyde for 30 min at room temperature.For adipogenesis assays, cells were incubated in 60% isopropa-nol for 5 min at room temperature, followed by a 15 min incu-bation in a filtered 3:2 dilution of Oil Red O stock solution (0.3%in isopropanol) and water. Cells were then washed with distilledwater and imaged with an inverted microscope (Zeiss AxioObserver). For osteogenesis experiments, fixed cells were stainedwith 2% Alizarin Red S in water (pH 4.2) for 45 min at roomtemperature, extensively washed with water and photographed.

For RNAseq experiments, cultured GLI1+ cells were starvedovernight in 0.5% FBS−containing GLI1+ medium, followed bystimulation with 100 nM SAG (Cayman) or vehicle (DMSO).After a 24-h incubation, RNA from three biological replicateswas extracted with the RNeasy Mini Kit (Qiagen) after directlysis of cells in the culture plate and processed for RNAseq. Bi-oinformatics analyses were performed as described above. Witha detection threshold set at ≥1 average FPKM in the condition withmost abundant expression, genes whose expression changed atleast twofold after SAG treatment (Benjamini–Hochberg correctedadj P < 0.05) were selected to carry out Gene Ontology analyseswith DAVID software (Huang et al., 2009). Ingenuity PathwayAnalysis (Qiagen) was performed with the same gene set butwithout FPKM or fold-change threshold. GSEAs (Subramanianet al., 2005) using the terms “immune response” (GO: 0006955),“cellular response to interferon-β” (GO: 0035458), and “cellularresponse to interferon-gamma” (GO: 0071346) were performed tocompare the transcriptomes of control and SAG-treated cells.

Mast cell survival assayConfluent GLI1+ cells grown in 48-well plates were starved in0.5% FBS-containing GLI1+ medium for 16 h, followed by stim-ulation with 100 nM SAG (Cayman) or vehicle (DMSO) for 24 h.As controls, we included DMSO- and SAG-containing wells



Dow



https://doi.org/10.1084/jem.20190730

without cells. RBL-2H3 cells (CRL-2256; ATCC) were labeledwith CellTracker Red CMTPX dye (C34552; Thermo Fisher Sci-entific) according to the manufacturer’s instructions, and25,000 cells/well were added on top of control or SAG-stimulated GLI1+ cells and on wells without GLI1+ cells, in theabsence or presence of SAG. After 24 h, cultures were stainedwith DAPI (1:2,000, catalog number D21490; Invitrogen) for2 min, and nonadherent cells were transferred to a new 48-wellplate and centrifuged at 400 g for 5 min. Cells were fixed with2% paraformaldehyde in PBS and images were obtained using anLSM510 confocal microscope using Zen software (Zeiss) in three1,270 × 1,270 µm fields per well. Cell death was calculated as thepercentage of CellTracker Red CMTPX dye–positive cells thatwere labeled with DAPI, using National Institutes of HealthImageJ software (version 1.49).

Melanocyte proliferation assay in conditioned media fromchoroidal GLI1+ cellsTo prepare conditioned media, GLI1+ cells were starved in 0.5%FBS-containing GLI1+ medium for 16 h, followed by stimulationwith 100 nM SAG or vehicle (DMSO) for 48 h. As controls, weincluded DMSO- and SAG-containing wells without cells. Mediawas collected, cleared by centrifugation and used immediately.Melan-a cells were cultured in RPMI media as previously de-scribed (Bennett et al., 1987) and seeded in triplicate (2,500cells/well) on black, clear-bottom 96-well plates in 100 µl cul-ture media. After 24 h, cells were starved for further 24 h in0.5% FBS-containing GLI1+ medium. Media was replaced withGLI1+ conditioned media prepared as described above. In par-allel, we included melan-a cells seeded at the same density andgrown in normal melan-a culture media that were representa-tive of maximum growth. Cell number was measured 48 h aftermedia change using the CyQUANT Direct Cell Proliferation As-say Kit (catalog number C35012; Thermo Fisher Scientific) andmeasuring fluorescence (480/535 nm excitation/emission) witha SpectraMax M2 Multi-Mode Microplane Reader (MolecularDevices). After background subtraction, relative cell numberwas calculated as the percentage of fluorescence intensity ofmelan-a cells cultured in their regular medium.

T cell proliferation assaysThe intracellular fluorescent dye CFSE (Invitrogen) was used todetermine cell division in responder cells as previously de-scribed (Lyons et al., 2013). Briefly, splenocytes were obtainedfrom BALB/c mice and washed using a 5-ml syringe to prepare asingle-cell suspension. Red blood cells were lysed with ammo-nium chloride lysing buffer (0.15 M NHCl, 1 mM KHCO, and0.1 mM EDTA) for 5 min at room temperature. Splenocytes werelabeled with 1 µM CFSE in PBS for 7 min at 37°C. The reactionwas stopped with ice-cold PBS supplemented with 10% FBS andwashed twice. Labeled cells were cultured either alone or in co-culture at different ratios with GLI1+ cells in the presence of 1 µgml−1 anti-CD3 monoclonal antibody (catalog number BE0001-1;Bio X Cell), which served as a cell activator. After 4 d of culture,responder nonadherent cells were harvested, stained with 0.5µg ml−1 PerCP- and PE-labeled anti-CD4 and anti-CD8a anti-bodies, respectively (catalog numbers 553052 and 553033; BD

PharMingen) and analyzed by flow cytometry (BD FACS CANTOII). CD4+ and CD8+ T cells were selected through gating andanalyzed for CFSE fluorescence intensity. Data analysis wasperformed using FlowJo software (version 8; TreeStar) to obtainthe division index (average number of cell divisions that a cell inthe original population has undergone) and the percentage ofcells from the original sample that divided.

NaIO3 treatment and optomotor response assaysVisual thresholds of Ihh+/+ and IhhiΔEC/iΔEC mice were measuredby evaluating optokinetic tracking in a virtual optokinetic sys-tem as described previously (Benedicto et al., 2017). Aftermeasuring baseline visual function, Ihh+/+ and IhhiΔEC/iΔEC micewere anesthetized and injected i.v. with a single dose of 15 mgkg−1 NaIO3 (Sigma) using a 30G needle. After 3 d, optomotorresponse assays were performed and mice were euthanized.Eyes were processed for RNA extraction from RPE/choroid andneural retina, and relative gene expression was assessed by real-time PCR (see above).

Statistical analysesAll graphs show individual data points and their average. Thenumber of biological replicates (n) is indicated in each figurelegend. Statistical significance was calculated using unpairedtwo-tailed t test or one-way ANOVA plus Bonferroni post hocanalysis (*, P < 0.05; **, P < 0.01; and ***, P < 0.001) as indicated.No statistical methodwas used to predetermine sample size or totest for normality and variance homogeneity. Data points >1.5interquartile ranges below the first quartile or above the thirdquartile were considered outliers and excluded from the anal-yses. All in vivo experiments were blinded and included litter-mate controls.

Ethical complianceAll animal protocols were reviewed and approved by the Insti-tutional Animal Care and Use Committee at Weill CornellMedicine.

Data availabilityAll data supporting the findings of this study are availablewithin the article and its supplemental information files or uponrequest. Bulk RNAseq data from mouse choroid ECs (Benedictoet al., 2017) were previously deposited in the Gene ExpressionOmnibus under series number GSE95835. RPE/choroid scRNA-seq and bulk RNAseq data (tissue-specific ECs, RPE/choroidfrom Ihh+/+ and IhhiΔEC/iΔECmice, and cultured GLI1+ cells ± SAG)have been deposited in the Gene Expression Omnibus underseries number GSE135167.

Online supplemental materialFig. S1 shows data regarding scRNAseq analyses, including thecell-sorting approach used to generate RPE/choroid single-cellsuspensions, superimposed tSNE plots from C57BL/6J andB6129PF1/J mouse strains, calculations for doublet estimation,and cluster-specific gene expression for C57BL/6J mice. Fig. S2shows the gating strategy used to analyze GLI1+ cells by flowcytometry. Fig. S3 shows loss of mast cells in Gli1-null mice,



Dow



https://doi.org/10.1084/jem.20190730

down-regulation of melanocyte markers in RPE/choroid tissueafter EC-specific Ihh deletion, and increased melanocyte prolif-eration in vitro in the presence of conditioned media fromSAG-stimulated GLI1+ cells. Fig. S4 shows the bioinformaticscharacterization of choroidal hematopoietic cells. Data S1shows complete datasets derived from scRNAseq and bulkRNAseq analyses.

AcknowledgmentsWe are thankful to the Genomics, Epigenomics, Flow Cytometry,and Visual Function core facilities at Weill Cornell Medicine andRyan Schreiner and Dena Almeida for their technical support.We thank Randi B. Silver for her insights on mast cell biologyand histochemistry. We thank Daniel Stephen (Memorial Sloan-Kettering Cancer Center, New York, NY) for supplying theGli1GFP/+ mice. We thank Verónica Labrador (Centro Nacional deInvestigaciones Cardiovasculares [CNIC] Microscopy and Dy-namic Imaging Unit) for her support with image analysis.

Funding for this study was provided by National Institutes ofHealth grants EY08538 and GM34107 (E. Rodriguez-Boulan);EY027038 (R.F. Mullins); 1R21CA224391-01A1 (J.H. Zippin); and1R01CA194547, 1U24CA210989, and P50CA211024 (O. Elemento);National Cancer Institute grant R01CA192176 and cancer centersupport grant P30 CA008748-48 (A.L. Joyner); Comunidad Au-tónoma de Madrid grant 2017-T1/BMD-5247 (I. Benedicto);Agencia Nacional Argentina de Promoción Cientıfica y Tecno-lógica grant PICT 2014-3687 and Fundación Sales (G.A. Rabinovich);a Pew Latin American Fellowship (G.L. Lehmann); Calder ResearchScholar Award Vitiligo/Pigment Cell Disorders (J.H. Zippin); StarrFoundation Tri-Institutional Stem Cell Initiative award 2013-028;NYSTEM contract C32596GG; and Research to Prevent Blindnessand Dyson Foundation departmental grants. The CNIC is supportedby the Instituto de Salud Carlos III, the Ministerio de Ciencia eInnovación, and the Pro CNIC Foundation and is a Severo OchoaCenter of Excellence (SEV-2015-0505).

Author contributions: Conceptualization, G.L. Lehmann, E.Rodriguez-Boulan, and I. Benedicto; methodology, G.L. Lehmannand I. Benedicto; investigation, G.L. Lehmann, C. Hanke-Gogokhia,Z. Salfati, M. Ginsberg, D.J. Nolan, A. Wojcinski, F. Ochoa, L.Panagis, S. Zeng, P.J. Zager, S. Ogura, J. Bang, J.P. Cerliani, S.P.Mendez-Huergo, T. Dalotto-Moreno, and I. Benedicto; softwareprogramming and bioinformatics, P.J. Zager, Y. Hu, R. Bareja, andO. Elemento; visualization, G.L. Lehmann, C. Hanke-Gogokhia,and I. Benedicto; writing (original draft), G.L. Lehmann and I.Benedicto; writing (review and editing), R.F. Mullins, G.A. Lutty,C. Hanke-Gogokhia, J.H. Zippin, C. Romano, G.A. Rabinovich, O.Elemento, A.L. Joyner, S. Rafii, E. Rodriguez-Boulan, and I.Benedicto; supervision, R.F. Mullins, G.A. Lutty, J.H. Zippin, C.Romano, G.A. Rabinovich, O. Elemento, A.L. Joyner, S. Rafii, E.Rodriguez-Boulan, and I. Benedicto; project administration, E.Rodriguez-Boulan and I. Benedicto.

Disclosures: Dr. Nolan reported personal fees from AngiocrineBioscience during the conduct of the study; in addition, Dr.Nolan had a patent number 9,944,897 issued "Angiocrine Bio-science." Dr. Panagis reported personal fees from Regeneron

outside the submitted work. Dr. Zippin reported grants fromPfizer and personal fees from Hoth outside the submitted work.Dr. Romano reported, "I am an employee of Regeneron Phar-maceuticals, however there are no products or services of thiscompany related to the work presented in this manuscript." Dr.Elemento reported "other" from Volastra Therapeutics, "other"from One Three Biotech, "other" from Freenome, and "other"from Owkin outside the submitted work. Dr. Rafii reported non-financial support from Angiocrine BioScience during the con-duct of the study. No other disclosures were reported.

Submitted: 23 April 2019Revised: 27 December 2019Accepted: 19 February 2020

ReferencesAhn, S., and A.L. Joyner. 2005. In vivo analysis of quiescent adult neural stem

cells responding to Sonic hedgehog. Nature. 437:894–897. https://doi.org/10.1038/nature03994

Alvarez, J.I., A. Dodelet-Devillers,H.Kebir, I. Ifergan, P.J. Fabre, S. Terouz,M. Sabbagh,K. Wosik, L. Bourbonnière, M. Bernard, et al. 2011. The Hedgehog pathwaypromotes blood-brain barrier integrity and CNS immune quiescence. Sci-ence. 334:1727–1731. https://doi.org/10.1126/science.1206936

Ambati, J., and B.J. Fowler. 2012. Mechanisms of age-related macular degen-eration. Neuron. 75:26–39. https://doi.org/10.1016/j.neuron.2012.06.018

Bai, C.B., W. Auerbach, J.S. Lee, D. Stephen, and A.L. Joyner. 2002. Gli2, butnot Gli1, is required for initial Shh signaling and ectopic activation ofthe Shh pathway. Development. 129:4753–4761.

Bai, C.B., D. Stephen, and A.L. Joyner. 2004. All mouse ventral spinal cordpatterning by hedgehog is Gli dependent and involves an activatorfunction of Gli3. Dev. Cell. 6:103–115. https://doi.org/10.1016/S1534-5807(03)00394-0

Benedicto, I., G.L. Lehmann, M. Ginsberg, D.J. Nolan, R. Bareja, O. Elemento,Z. Salfati, N.M. Alam, G.T. Prusky, P. Llanos, et al. 2017. Concertedregulation of retinal pigment epithelium basement membrane andbarrier function by angiocrine factors. Nat. Commun. 8:15374. https://doi.org/10.1038/ncomms15374

Bennett, D.C., P.J. Cooper, and I.R. Hart. 1987. A line of non-tumorigenicmouse melanocytes, syngeneic with the B16 melanoma and requiringa tumour promoter for growth. Int. J. Cancer. 39:414–418. https://doi.org/10.1002/ijc.2910390324

Berg, S., D. Kutra, T. Kroeger, C.N. Straehle, B.X. Kausler, C. Haubold, M.Schiegg, J. Ales, T. Beier, M. Rudy, et al. 2019. ilastik: interactive ma-chine learning for (bio)image analysis. Nat. Methods. 16:1226–1232.https://doi.org/10.1038/s41592-019-0582-9

Bhutto, I., and G. Lutty. 2012. Understanding age-related macular degener-ation (AMD): relationships between the photoreceptor/retinal pigmentepithelium/Bruch’s membrane/choriocapillaris complex. Mol. AspectsMed. 33:295–317. https://doi.org/10.1016/j.mam.2012.04.005

Bhutto, I.A., D.S. McLeod, T. Jing, J.S. Sunness, J.M. Seddon, and G.A. Lutty.2016. Increased choroidal mast cells and their degranulation in age-related macular degeneration. Br. J. Ophthalmol. 100:720–726. https://doi.org/10.1136/bjophthalmol-2015-308290

Biswas, S.K., and A. Mantovani. 2010. Macrophage plasticity and interactionwith lymphocyte subsets: cancer as a paradigm. Nat. Immunol. 11:889–896. https://doi.org/10.1038/ni.1937

Brownell, I., E. Guevara, C.B. Bai, C.A. Loomis, and A.L. Joyner. 2011. Nerve-derived sonic hedgehog defines a niche for hair follicle stem cells ca-pable of becoming epidermal stem cells. Cell Stem Cell. 8:552–565.https://doi.org/10.1016/j.stem.2011.02.021

Butler, A., P. Hoffman, P. Smibert, E. Papalexi, and R. Satija. 2018. Integratingsingle-cell transcriptomic data across different conditions, technologies,and species. Nat. Biotechnol. 36:411–420. https://doi.org/10.1038/nbt.4096

Caughey, G.H. 2011. Mast cell proteases as protective and inflammatorymediators. Adv. Exp. Med. Biol. 716:212–234. https://doi.org/10.1007/978-1-4419-9533-9_12

Chang, B., N.L. Hawes, R.E. Hurd, M.T. Davisson, S. Nusinowitz, and J.R.Heckenlively. 2002. Retinal degeneration mutants in the mouse. VisionRes. 42:517–525. https://doi.org/10.1016/S0042-6989(01)00146-8



Dow



https://doi.org/10.1038/nature03994


https://doi.org/10.1126/science.1206936

https://doi.org/10.1016/j.neuron.2012.06.018

https://doi.org/10.1016/S1534-5807(03)00394-0

https://doi.org/10.1016/S1534-5807(03)00394-0

https://doi.org/10.1038/ncomms15374

https://doi.org/10.1038/ncomms15374

https://doi.org/10.1002/ijc.2910390324

https://doi.org/10.1002/ijc.2910390324

https://doi.org/10.1038/s41592-019-0582-9

https://doi.org/10.1016/j.mam.2012.04.005

https://doi.org/10.1136/bjophthalmol-2015-308290

https://doi.org/10.1136/bjophthalmol-2015-308290

https://doi.org/10.1038/ni.1937

https://doi.org/10.1016/j.stem.2011.02.021

https://doi.org/10.1038/nbt.4096

https://doi.org/10.1007/978-1-4419-9533-9_12

https://doi.org/10.1007/978-1-4419-9533-9_12

https://doi.org/10.1016/S0042-6989(01)00146-8

https://doi.org/10.1084/jem.20190730

Chen, J.K., J. Taipale, K.E. Young, T. Maiti, and P.A. Beachy. 2002. Smallmolecule modulation of Smoothened activity. Proc. Natl. Acad. Sci. USA.99:14071–14076. https://doi.org/10.1073/pnas.182542899

Dakubo, G.D., Y.P. Wang, C. Mazerolle, K. Campsall, A.P. McMahon, and V.A.Wallace. 2003. Retinal ganglion cell-derived sonic hedgehog signaling isrequired for optic disc and stalk neuroepithelial cell development. De-velopment. 130:2967–2980. https://doi.org/10.1242/dev.00515

Dakubo, G.D., C. Mazerolle, M. Furimsky, C. Yu, B. St-Jacques, A.P. McMa-hon, and V.A. Wallace. 2008. Indian hedgehog signaling from endo-thelial cells is required for sclera and retinal pigment epitheliumdevelopment in the mouse eye. Dev. Biol. 320:242–255. https://doi.org/10.1016/j.ydbio.2008.05.528

Dery, O., C.U. Corvera, M. Steinhoff, and N.W. Bunnett. 1998. Proteinase-activated receptors: novel mechanisms of signaling by serine proteases.Am. J. Physiol. 274:C1429–C1452. https://doi.org/10.1152/ajpcell.1998.274.6.C1429

Dobin, A., C.A. Davis, F. Schlesinger, J. Drenkow, C. Zaleski, S. Jha, P. Batut, M.Chaisson, and T.R. Gingeras. 2013. STAR: ultrafast universal RNA-seqaligner. Bioinformatics. 29:15–21. https://doi.org/10.1093/bioinformatics/bts635

Dwyer, D.F., N.A. Barrett, and K.F. Austen. Immunological Genome ProjectConsortium. 2016. Expression profiling of constitutive mast cells re-veals a unique identity within the immune system. Nat. Immunol. 17:878–887. https://doi.org/10.1038/ni.3445

Galli, S.J., M. Grimbaldeston, and M. Tsai. 2008. Immunomodulatory mastcells: negative, as well as positive, regulators of immunity. Nat. Rev.Immunol. 8:478–486. https://doi.org/10.1038/nri2327

Han, X., R. Wang, Y. Zhou, L. Fei, H. Sun, S. Lai, A. Saadatpour, Z. Zhou, H.Chen, F. Ye, et al. 2018. Mapping the Mouse Cell Atlas by Microwell-Seq. Cell. 172:1091–1107.e17. https://doi.org/10.1016/j.cell.2018.02.001

Hart, P.H., M.A. Grimbaldeston, G.J. Swift, A. Jaksic, F.P. Noonan, and J.J.Finlay-Jones. 1998. Dermal mast cells determine susceptibility to ul-traviolet B-induced systemic suppression of contact hypersensitivityresponses in mice. J. Exp. Med. 187:2045–2053. https://doi.org/10.1084/jem.187.12.2045

Heng, T.S., and M.W. Painter. Immunological Genome Project Consortium.2008. The Immunological Genome Project: networks of gene expres-sion in immune cells. Nat. Immunol. 9:1091–1094. https://doi.org/10.1038/ni1008-1091

Huang, W., B.T. Sherman, and R.A. Lempicki. 2009. Systematic and inte-grative analysis of large gene lists using DAVID bioinformatics re-sources. Nat. Protoc. 4:44–57. https://doi.org/10.1038/nprot.2008.211

Kannan, R., and D.R. Hinton. 2014. Sodium iodate induced retinal degener-ation: new insights from an old model. Neural Regen. Res. 9:2044–2045.https://doi.org/10.4103/1673-5374.147927

Kramann, R., R.K. Schneider, D.P. DiRocco, F. Machado, S. Fleig, P.A. Bondzie,J.M. Henderson, B.L. Ebert, and B.D. Humphreys. 2015. PerivascularGli1+ progenitors are key contributors to injury-induced organ fibrosis.Cell Stem Cell. 16:51–66. https://doi.org/10.1016/j.stem.2014.11.004

Lee, J.J., M.E. Rothenberg, E.S. Seeley, B. Zimdahl, S. Kawano, W.J. Lu, K. Shin,T. Sakata-Kato, J.K. Chen, M. Diehn, et al. 2016. Control of inflammationby stromal Hedgehog pathway activation restrains colitis. Proc. Natl.Acad. Sci. USA. 113:E7545–E7553. https://doi.org/10.1073/pnas.1616447113

Lees, C.W.,W.J. Zacharias, M. Tremelling, C.L. Noble, E.R. Nimmo, A. Tenesa,J. Cornelius, L. Torkvist, J. Kao, S. Farrington, et al. 2008. Analysis ofgermline GLI1 variation implicates hedgehog signalling in the regula-tion of intestinal inflammatory pathways. PLoS Med. 5:e239. https://doi.org/10.1371/journal.pmed.0050239

Li, H., B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G.Abecasis, and R. Durbin. 1000 Genome Project Data Processing Subgroup.2009. The Sequence Alignment/Map format and SAMtools. Bioinformatics.25:2078–2079. https://doi.org/10.1093/bioinformatics/btp352

Love, M.I., W. Huber, and S. Anders. 2014. Moderated estimation of foldchange and dispersion for RNA-seq data with DESeq2. Genome Biol. 15:550. https://doi.org/10.1186/s13059-014-0550-8

Lu, L.F., E.F. Lind, D.C. Gondek, K.A. Bennett, M.W. Gleeson, K. Pino-Lagos,Z.A. Scott, A.J. Coyle, J.L. Reed, J. Van Snick, et al. 2006. Mast cells areessential intermediaries in regulatory T-cell tolerance. Nature. 442:997–1002. https://doi.org/10.1038/nature05010

Lutty, G.A., T. Hasegawa, T. Baba, R. Grebe, I. Bhutto, and D.S. McLeod. 2010.Development of the human choriocapillaris. Eye (Lond.). 24:408–415.https://doi.org/10.1038/eye.2009.318

Lyons, A.B., S.J. Blake, and K.V. Doherty. 2013. Flow cytometric analysis ofcell division by dilution of CFSE and related dyes. Curr Protoc Cytom.Chapter 9:Unit9.11.

Macosko, E.Z., A. Basu, R. Satija, J. Nemesh, K. Shekhar, M. Goldman, I.Tirosh, A.R. Bialas, N. Kamitaki, E.M. Martersteck, et al. 2015. HighlyParallel Genome-wide Expression Profiling of Individual Cells UsingNanoliter Droplets. Cell. 161:1202–1214. https://doi.org/10.1016/j.cell.2015.05.002

May, C.A. 1999. Mast cell heterogeneity in the human uvea. Histochem. CellBiol. 112:381–386. https://doi.org/10.1007/s004180050420

McDavid, A., G. Finak, P.K. Chattopadyay, M. Dominguez, L. Lamoreaux, S.S.Ma, M. Roederer, and R. Gottardo. 2013. Data exploration, quality controland testing in single-cell qPCR-based gene expression experiments. Bio-informatics. 29:461–467. https://doi.org/10.1093/bioinformatics/bts714

McLeod, D.S., I. Bhutto, M.M. Edwards, M. Gedam, R. Baldeosingh, and G.A.Lutty. 2017. Mast Cell-Derived Tryptase in Geographic Atrophy. Invest.Ophthalmol. Vis. Sci. 58:5887–5896. https://doi.org/10.1167/iovs.17-22989

McMenamin, P.G. 1997. The distribution of immune cells in the uveal tract of thenormal eye. Eye (Lond.). 11:183–193. https://doi.org/10.1038/eye.1997.49

McMenamin, P.G., and E. Polla. 2013. Mast cells are present in the choroid ofthe normal eye in most vertebrate classes. Vet. Ophthalmol. 16(Suppl 1):73–78. https://doi.org/10.1111/vop.12035

Merad, M., P. Sathe, J. Helft, J. Miller, and A. Mortha. 2013. The dendritic celllineage: ontogeny and function of dendritic cells and their subsets in thesteady state and the inflamed setting. Annu. Rev. Immunol. 31:563–604.https://doi.org/10.1146/annurev-immunol-020711-074950

Merchant, A., G. Joseph, Q. Wang, S. Brennan, and W. Matsui. 2010. Gli1regulates the proliferation and differentiation of HSCs and myeloidprogenitors. Blood. 115:2391–2396. https://doi.org/10.1182/blood-2009-09-241703

Mullins, R.F., M.N. Johnson, E.A. Faidley, J.M. Skeie, and J. Huang. 2011.Choriocapillaris vascular dropout related to density of drusen in humaneyes with early age-related macular degeneration. Invest. Ophthalmol.Vis. Sci. 52:1606–1612. https://doi.org/10.1167/iovs.10-6476

Murray, P.J. 2017. Macrophage Polarization. Annu. Rev. Physiol. 79:541–566.https://doi.org/10.1146/annurev-physiol-022516-034339

Nazari, H., L. Zhang, D. Zhu, G.J. Chader, P. Falabella, F. Stefanini, T. Row-land, D.O. Clegg, A.H. Kashani, D.R. Hinton, and M.S. Humayun. 2015.Stem cell based therapies for age-related macular degeneration: Thepromises and the challenges. Prog. Retin. Eye Res. 48:1–39. https://doi.org/10.1016/j.preteyeres.2015.06.004

Newman, A.M., N.B. Gallo, L.S. Hancox, N.J. Miller, C.M. Radeke, M.A. Ma-loney, J.B. Cooper, G.S. Hageman, D.H. Anderson, L.V. Johnson, andM.J.Radeke. 2012. Systems-level analysis of age-related macular degenera-tion reveals global biomarkers and phenotype-specific functional net-works. Genome Med. 4:16. https://doi.org/10.1186/gm315

Nickla, D.L., and J. Wallman. 2010. The multifunctional choroid. Prog. Retin.Eye Res. 29:144–168. https://doi.org/10.1016/j.preteyeres.2009.12.002

Park, H.L., C. Bai, K.A. Platt,M.P.Matise, A. Beeghly, C.C. Hui,M. Nakashima,and A.L. Joyner. 2000. Mouse Gli1 mutants are viable but have defectsin SHH signaling in combinationwith a Gli2 mutation. Development. 127:1593–1605.

Passante, E., and N. Frankish. 2009. The RBL-2H3 cell line: its provenanceand suitability as a model for the mast cell. Inflamm. Res. 58:737–745.https://doi.org/10.1007/s00011-009-0074-y

Petrova, R., and A.L. Joyner. 2014. Roles for Hedgehog signaling in adult organhomeostasis and repair. Development. 141:3445–3457. https://doi.org/10.1242/dev.083691

Rafii, S., J.M. Butler, and B.S. Ding. 2016. Angiocrine functions of organ-specific endothelial cells. Nature. 529:316–325. https://doi.org/10.1038/nature17040

Seddon, J.M., D.S. McLeod, I.A. Bhutto, M.B. Villalonga, R.E. Silver, A.S.Wenick, M.M. Edwards, and G.A. Lutty. 2016. Histopathological In-sights Into Choroidal Vascular Loss in Clinically Documented Cases ofAge-Related Macular Degeneration. JAMA Ophthalmol. 134:1272–1280.https://doi.org/10.1001/jamaophthalmol.2016.3519

Smith, J.R., P.H. Hart, D.J. Coster, and K.A. Williams. 1998. Uveal mast cellsare not required for rodent uveitis. Ophthalmic Res. 30:388–393. https://doi.org/10.1159/000055500

Songstad, A.E., L.A. Wiley, K. Duong, E. Kaalberg, M.J. Flamme-Wiese, C.M.Cranston, M.J. Riker, D. Levasseur, E.M. Stone, R.F. Mullins, and B.A.Tucker. 2015. Generating iPSC-Derived Choroidal Endothelial Cells toStudy Age-Related Macular Degeneration. Invest. Ophthalmol. Vis. Sci.56:8258–8267. https://doi.org/10.1167/iovs.15-17073

Songstad, A.E., K.S. Worthington, K.R. Chirco, J.C. Giacalone, S.S. Whitmore,K.R. Anfinson, D. Ochoa, C.M. Cranston, M.J. Riker, M. Neiman, et al.2017. Connective Tissue Growth Factor Promotes Efficient Generationof Human Induced Pluripotent Stem Cell-Derived Choroidal



Dow



https://doi.org/10.1073/pnas.182542899

https://doi.org/10.1242/dev.00515

https://doi.org/10.1016/j.ydbio.2008.05.528

https://doi.org/10.1016/j.ydbio.2008.05.528

https://doi.org/10.1152/ajpcell.1998.274.6.C1429

https://doi.org/10.1152/ajpcell.1998.274.6.C1429

https://doi.org/10.1093/bioinformatics/bts635


https://doi.org/10.1038/ni.3445

https://doi.org/10.1038/nri2327

https://doi.org/10.1016/j.cell.2018.02.001

https://doi.org/10.1084/jem.187.12.2045

https://doi.org/10.1084/jem.187.12.2045

https://doi.org/10.1038/ni1008-1091

https://doi.org/10.1038/ni1008-1091

https://doi.org/10.1038/nprot.2008.211

https://doi.org/10.4103/1673-5374.147927



https://doi.org/10.1371/journal.pmed.0050239

https://doi.org/10.1371/journal.pmed.0050239

https://doi.org/10.1093/bioinformatics/btp352

https://doi.org/10.1186/s13059-014-0550-8


https://doi.org/10.1038/eye.2009.318



https://doi.org/10.1007/s004180050420


https://doi.org/10.1167/iovs.17-22989

https://doi.org/10.1038/eye.1997.49

https://doi.org/10.1111/vop.12035

https://doi.org/10.1146/annurev-immunol-020711-074950

https://doi.org/10.1182/blood-2009-09-241703

https://doi.org/10.1182/blood-2009-09-241703


https://doi.org/10.1146/annurev-physiol-022516-034339

https://doi.org/10.1016/j.preteyeres.2015.06.004


https://doi.org/10.1186/gm315


https://doi.org/10.1007/s00011-009-0074-y





https://doi.org/10.1001/jamaophthalmol.2016.3519

https://doi.org/10.1159/000055500

https://doi.org/10.1159/000055500


https://doi.org/10.1084/jem.20190730

Endothelium. Stem Cells Transl. Med. 6:1533–1546. https://doi.org/10.1002/sctm.16-0399

St-Jacques, B., M. Hammerschmidt, and A.P. McMahon. 1999. Indianhedgehog signaling regulates proliferation and differentiation ofchondrocytes and is essential for bone formation. Genes Dev. 13:2072–2086. https://doi.org/10.1101/gad.13.16.2072

Stan, R.V., D. Tse, S.J. Deharvengt, N.C. Smits, Y. Xu, M.R. Luciano, C.L.McGarry, M. Buitendijk, K.V. Nemani, R. Elgueta, et al. 2012. The dia-phragms of fenestrated endothelia: gatekeepers of vascular permea-bility and blood composition. Dev. Cell. 23:1203–1218. https://doi.org/10.1016/j.devcel.2012.11.003

Steinhoff, M., N. Vergnolle, S.H. Young, M. Tognetto, S. Amadesi, H.S. Ennes,M. Trevisani, M.D. Hollenberg, J.L. Wallace, G.H. Caughey, et al. 2000.Agonists of proteinase-activated receptor 2 induce inflammation by aneurogenic mechanism. Nat. Med. 6:151–158. https://doi.org/10.1038/72247

Strauss, O. 2005. The retinal pigment epithelium in visual function. Physiol.Rev. 85:845–881. https://doi.org/10.1152/physrev.00021.2004

Subramanian, A., P. Tamayo, V.K. Mootha, S. Mukherjee, B.L. Ebert, M.A. Gil-lette, A. Paulovich, S.L. Pomeroy, T.R. Golub, E.S. Lander, and J.P. Mesirov.2005. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA.102:15545–15550. https://doi.org/10.1073/pnas.0506580102

Tharp, M.D., L.L. Seelig Jr., R.E. Tigelaar, and P.R. Bergstresser. 1985. Con-jugated avidin binds to mast cell granules. J. Histochem. Cytochem. 33:27–32. https://doi.org/10.1177/33.1.2578142

Trapnell, C., A. Roberts, L. Goff, G. Pertea, D. Kim, D.R. Kelley, H. Pimentel,S.L. Salzberg, J.L. Rinn, and L. Pachter. 2012. Differential gene andtranscript expression analysis of RNA-seq experiments with TopHatand Cufflinks. Nat. Protoc. 7:562–578. https://doi.org/10.1038/nprot.2012.016

Tsai, M., M. Grimbaldeston, and S.J. Galli. 2011. Mast cells and immunoreg-ulation/immunomodulation. Adv. Exp. Med. Biol. 716:186–211. https://doi.org/10.1007/978-1-4419-9533-9_11

Uccelli, A., L. Moretta, and V. Pistoia. 2008.Mesenchymal stem cells in healthand disease. Nat. Rev. Immunol. 8:726–736. https://doi.org/10.1038/nri2395

Van der Maaten, L.H.G. 2008. Visualizing data using t-SNE. J. Mach. Learn.Res. 9:2579–2605.

van Dop, W.A., J. Heijmans, N.V. Büller, S.A. Snoek, S.L. Rosekrans, E.A.Wassenberg, M.A. van den BerghWeerman, B. Lanske, A.R. Clarke, D.J.Winton, et al. 2010. Loss of Indian Hedgehog activates multiple aspectsof a wound healing response in the mouse intestine. Gastroenterology.

139:1665–1676: 1676.e1–1676.e10. https://doi.org/10.1053/j.gastro.2010.07.045

Veerappan, A., A.C. Reid, R. Estephan, N. O’Connor, M. Thadani-Mulero, M.Salazar-Rodriguez, R. Levi, and R.B. Silver. 2008. Mast cell renin and alocal renin-angiotensin system in the airway: role in bronchoconstric-tion. Proc. Natl. Acad. Sci. USA. 105:1315–1320. https://doi.org/10.1073/pnas.0709739105

Wallace, V.A., and M.C. Raff. 1999. A role for Sonic hedgehog in axon-to-astrocyte signalling in the rodent optic nerve. Development. 126:2901–2909.

Wang, Y., M. Nakayama, M.E. Pitulescu, T.S. Schmidt, M.L. Bochenek, A.Sakakibara, S. Adams, A. Davy, U. Deutsch, U. Lüthi, et al. 2010. Ephrin-B2 controls VEGF-induced angiogenesis and lymphangiogenesis. Na-ture. 465:483–486. https://doi.org/10.1038/nature09002

Westendorp, B.F., N. Buller, O.N. Karpus, W.A. van Dop, J. Koster, R. Ver-steeg, P.J. Koelink, C.Y. Snel, S. Meisner, J. Roelofs, et al. 2018. IndianHedgehog Suppresses a Stromal Cell-Driven Intestinal Immune Re-sponse. Cell. Mol. Gastroenterol. Hepatol. 5:67–82.e1. https://doi.org/10.1016/j.jcmgh.2017.08.004

Whitmore, S.S., E.H. Sohn, K.R. Chirco, A.V. Drack, E.M. Stone, B.A. Tucker,and R.F. Mullins. 2015. Complement activation and choriocapillaris lossin early AMD: implications for pathophysiology and therapy. Prog.Retin. Eye Res. 45:1–29. https://doi.org/10.1016/j.preteyeres.2014.11.005

Wong, W.L., X. Su, X. Li, C.M. Cheung, R. Klein, C.Y. Cheng, and T.Y. Wong.2014. Global prevalence of age-related macular degeneration and dis-ease burden projection for 2020 and 2040: a systematic review andmeta-analysis. Lancet Glob. Health. 2:e106–e116. https://doi.org/10.1016/S2214-109X(13)70145-1

Wu, F., Y. Zhang, B. Sun, A.P. McMahon, and Y. Wang. 2017. HedgehogSignaling: From Basic Biology to Cancer Therapy. Cell Chem. Biol. 24:252–280. https://doi.org/10.1016/j.chembiol.2017.02.010

Zacharias, W.J., X. Li, B.B. Madison, K. Kretovich, J.Y. Kao, J.L. Merchant, andD.L. Gumucio. 2010. Hedgehog is an anti-inflammatory epithelial signalfor the intestinal lamina propria. Gastroenterology. 138:2368–2377:2377.e1–2377.e4. https://doi.org/10.1053/j.gastro.2010.02.057

Zeng, S., J. Hernandez, and R.F. Mullins. 2012. Effects of antioxidant com-ponents of AREDS vitamins and zinc ions on endothelial cell activation:implications for macular degeneration. Invest. Ophthalmol. Vis. Sci. 53:1041–1047. https://doi.org/10.1167/iovs.11-8531

Zhao, H., J. Feng, K. Seidel, S. Shi, O. Klein, P. Sharpe, and Y. Chai. 2014.Secretion of shh by a neurovascular bundle niche supports mesen-chymal stem cell homeostasis in the adult mouse incisor. Cell Stem Cell.14:160–173. https://doi.org/10.1016/j.stem.2013.12.013



Dow



https://doi.org/10.1002/sctm.16-0399

https://doi.org/10.1002/sctm.16-0399

https://doi.org/10.1101/gad.13.16.2072

https://doi.org/10.1016/j.devcel.2012.11.003

https://doi.org/10.1016/j.devcel.2012.11.003

https://doi.org/10.1038/72247

https://doi.org/10.1038/72247

https://doi.org/10.1152/physrev.00021.2004


https://doi.org/10.1177/33.1.2578142



https://doi.org/10.1007/978-1-4419-9533-9_11

https://doi.org/10.1007/978-1-4419-9533-9_11



https://doi.org/10.1053/j.gastro.2010.07.045





https://doi.org/10.1016/j.jcmgh.2017.08.004

https://doi.org/10.1016/j.jcmgh.2017.08.004


https://doi.org/10.1016/S2214-109X(13)70145-1

https://doi.org/10.1016/S2214-109X(13)70145-1

https://doi.org/10.1016/j.chembiol.2017.02.010




https://doi.org/10.1084/jem.20190730

Supplemental material

Lehmann et al. Journal of Experimental Medicine S1


Dow



https://doi.org/10.1084/jem.20190730

Figure S1. scRNAseq of RPE/choroid tissue. (a) Cell sorting to prepare RPE/choroid single-cell suspensions suitable for scRNAseq. Experiments wereperformed with the strains C57BL/6J (top panels) and B6129PF1/J (bottom panels). For each strain, RPE/choroid tissue from one male and one female eye waspooled, digested, and stained with Hoechst and TO-PRO-3 to label nucleated and dead cells, respectively. Unstained samples were used as controls to set thegating parameters. Pilot experiments showed two different Hoechst+/TO-PRO-3− populations with dim (blue dots) and high (red dots) Hoechst staining.Hoechstdim events presented very low forward scatter (FSC; right panels), suggesting that this population could be debris. To assess the cellular content of bothpopulations, we sorted Hoechstdim/TO-PRO-3− and Hoechsthigh/TO-PRO-3− events independently and observed them under the microscope. The Hoechstdim

population contained cell debris (dashed yellow squares) but also many cells that were mainly pigmented (black arrows). Hence, our results suggest that cellpigment present in RPE and melanocytes may decrease the apparent FSC of both cell types. The Hoechsthigh population (red) was virtually free of debris andmainly constituted by nonpigmented cells (white arrows). To avoid loss of pigmented cells, both Hoechstdim/TO-PRO-3− and Hoechsthigh/TO-PRO-3− pop-ulations were sorted together in the same collecting tube and used for scRNAseq assays. Scale bar, 50 µm. (b) Estimation of doublet content in scRNAseqassays (see calculations in Materials and methods). (c) Principal-component analysis of 7,723 single-cell expression profiles obtained from a mix of male (n =1 per strain) and female (n = 1 per strain) RPE/choroid tissue. Data are shown in two dimensions using tSNE after CCA batch correction. Black dots, 3,996 cellsfrom C57BL/6J mice; red dots, 3,727 cells from B6129PF1/J mice. (d) Identification of cell types in RPE/choroid tissue from C57BL/6J mice according to theaverage expression of known markers in each cluster. Cluster number colors are the same as in Fig. 1 (b and c). Relative gene expression among cell types wascalculated using the average normalized UMIs in each cluster and represented as the percentage of the cluster with maximum expression. White, 0%;red, 100%.



Dow



https://doi.org/10.1084/jem.20190730

Figure S2. Gating strategy to quantify and characterize GLI1+ cells from Gli1GFP/+ mice by flow cytometry. (a and b) Analysis of GFP+ cells from RPE/choroid (a) and lung (b). Tissue from Gli1+/+ mice was used as negative control. FSC-H, forward scatter height; SSC-A, side scatter area; SSC-H, side scatterheight; SSC-W, side scatter width.



Dow



https://doi.org/10.1084/jem.20190730

Figure S3. Effects of HH signaling on choroidal mast cells and melanocyte proliferation. (a) Real-time PCR showing the expression of mast cell markersin RPE/choroid tissue from Gli1+/+ and Gli1GFP/GFP mice (black and green dots, respectively). Individual dots correspond to different animals, and red barsrepresent average values. Relative gene expression is presented as the percentage of Gli1+/+ mice (n = 4, t test). *, P < 0.05; **, P < 0.01. (b) Representativeavidin staining assay to localize choroidal mast cells using flatmounts from albino Gli1+/+ and Gli1GFP/GFPmice. Panels show binary images used for segmentation,and dashed lines delineate flatmount contours. Scale bar, 500 µm. (c) Quantification of avidin staining assays shown in panel b. Results are expressed asnumber of mast cells normalized by flatmount area and presented as relative units (Gli1+/+, n = 3; Gli1GFP/GFP, n = 4; t test). **, P < 0.01. (d) Differential ex-pression of melanocyte-enriched genes in RPE/choroid from Ihh+/+ and IhhiΔEC/iΔEC mice as determined by bulk RNAseq (n = 3, Benjamini–Hochberg correctedadj P < 0.05). (e) Serum-starved melan-a cells were grown in culture medium alone, culture medium supplemented with 100 nM SAG, or conditioned media(CM) from unstimulated or 100 nM SAG–treated GLI1+ cells. All conditions included 0.5% FBS. After 48 h, cell number was assessed and represented as thepercentage of cells grown in normal melan-a culture medium with 10% FBS. Individual dots correspond to three independent experiments performed inbiological triplicates, and red bars represent average values (n = 3, ANOVA + Bonferroni test). *, P < 0.05.



Dow



https://doi.org/10.1084/jem.20190730

Figure S4. Classification of choroidal hematopoietic cell subclusters. (a) Gene lists from Data S1 m were analyzed at the Immunological Genome Projectwebsite (http://www.immgen.org) using My GeneSet function (V1 datasets for subclusters H1-4, V2 datasets for subcluster H5). (b) Expression of CD103−

CD11b+ and CD103+ CD11b− cDC markers (based on Merad et al., 2013) by H1 and H4 subclusters. Average normalized UMIs for each gene and subcluster wereextracted from Data S1 l.



Dow



http://www.immgen.org

https://doi.org/10.1084/jem.20190730

A supplemental dataset is available online that provides information regarding all scRNAseq and bulk RNAseq differentialexpression analyses used in the study.



Dow



https://doi.org/10.1084/jem.20190730

article single-cell profiling reveals an endothelium-mediated … · 2019. 7. 30. · article...

Documents